Keyword: machine learning
2018 (17)
Multiobjective Optimization for Stiffness and Position Control in a Soft Robot Arm Module. Ansari, Y., Manti, M., Falotico, E., Cianchetti, M., & Laschi, C. IEEE Robotics and Automation Letters, 3(1):108–115, January, 2018.
doi abstract bibtex
doi abstract bibtex
The central concept of this letter is to develop an assistive manipulator that can automate the bathing task for elderly citizens. We propose to exploit principles of soft robotic technologies to design and control a compliant system to ensure safe human–robot interaction, a primary requirement for the task. The overall system is intended to be modular with a proximal segment that provides structural integrity to overcome gravitational challenges and a distal segment to perform the main bathing activities. The focus of this letter is on the design and control of the latter module. The design comprises of alternating tendons and pneumatics in a radial arrangement, which enables elongation, contraction, and omnidirectional bending. Additionally, a synergetic coactivation of cables and tendons in a given configuration allows for stiffness modulation, which is necessary to facilitate washing and scrubbing. The novelty of the work is twofold: 1) Three base cases of antagonistic actuation are identified that enable stiffness variation. Each category is then experimentally characterized by the application of an external force that imposes a linear displacement at the tip in both axial and lateral directions. 2) The development of a novel algorithm based on cooperative multiagent reinforcement learning that simultaneously optimizes stiffness and position. The results highlight the effectiveness of the design and control to contribute toward the development of the assistive device.
@article{ansari_multiobjective_2018,
title = {Multiobjective {Optimization} for {Stiffness} and {Position} {Control} in a {Soft} {Robot} {Arm} {Module}},
volume = {3},
doi = {10.1109/LRA.2017.2734247},
abstract = {The central concept of this letter is to develop an assistive manipulator that can automate the bathing task for elderly citizens. We propose to exploit principles of soft robotic technologies to design and control a compliant system to ensure safe human–robot interaction, a primary requirement for the task. The overall system is intended to be modular with a proximal segment that provides structural integrity to overcome gravitational challenges and a distal segment to perform the main bathing activities. The focus of this letter is on the design and control of the latter module. The design comprises of alternating tendons and pneumatics in a radial arrangement, which enables elongation, contraction, and omnidirectional bending. Additionally, a synergetic coactivation of cables and tendons in a given configuration allows for stiffness modulation, which is necessary to facilitate washing and scrubbing. The novelty of the work is twofold: 1) Three base cases of antagonistic actuation are identified that enable stiffness variation. Each category is then experimentally characterized by the application of an external force that imposes a linear displacement at the tip in both axial and lateral directions. 2) The development of a novel algorithm based on cooperative multiagent reinforcement learning that simultaneously optimizes stiffness and position. The results highlight the effectiveness of the design and control to contribute toward the development of the assistive device.},
number = {1},
journal = {IEEE Robotics and Automation Letters},
author = {Ansari, Y. and Manti, M. and Falotico, E. and Cianchetti, M. and Laschi, C.},
month = jan,
year = {2018},
keywords = {Assistive robotics, Force, Robot control, Tendons, actuators, learning (artificial intelligence), machine learning, manipulators, soft robotics},
pages = {108--115}
}Gaze and the Control of Foot Placement When Walking in Natural Terrain. Matthis, J. S., Yates, J. L., & Hayhoe, M. M. Current Biology, April, 2018.
![Gaze and the Control of Foot Placement When Walking in Natural Terrain [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi bibtex
Paper doi bibtex @article{matthis_gaze_2018,
title = {Gaze and the {Control} of {Foot} {Placement} {When} {Walking} in {Natural} {Terrain}},
volume = {0},
issn = {0960-9822},
url = {http://www.cell.com/current-biology/abstract/S0960-9822(18)30309-9},
doi = {10.1016/j.cub.2018.03.008},
language = {English},
number = {0},
urldate = {2018-04-13},
journal = {Current Biology},
author = {Matthis, Jonathan Samir and Yates, Jacob L. and Hayhoe, Mary M.},
month = apr,
year = {2018},
keywords = {eye movements, foot placement, gaze, locomotion, real-world, rough terrain, walking}
}Sleep Duration and Physical Activity Profiles Associated With Self-Reported Stroke in the United States: Application of Bayesian Belief Network Modeling Techniques. Seixas, A., A., Henclewood, D., A., Williams, S., K., Jagannathan, R., Ramos, A., Zizi, F., & Jean-Louis, G. Frontiers in Neurology, 9:534, Frontiers, 7, 2018.
Website doi abstract bibtex
Website doi abstract bibtex Introduction: Physical activity (PA) and sleep are associated with cerebrovascular disease and events like stroke. Though the interrelationships between PA, sleep, and other stroke risk factors have been studied, we are unclear about the associations of different types, frequency and duration of PA, sleep behavioral patterns (short, average and long sleep durations), within the context of stroke-related clinical, behavioral, and socio-demographic risk factors. The current study utilized Bayesian Belief Network analysis, a type of machine learning analysis, to develop profiles of physical activity (duration, intensity, and frequency) and sleep duration associated with or no history of stroke, given the influence of multiple stroke predictors and correlates. Such a model allowed us to develop a predictive classification model of stroke which can be used in post-stroke risk stratification and developing targeted stroke rehabilitation care based on an individual’s profile. Method: Analysis was based on the 2004-2013 National Health Interview Survey (n=288,888). Bayesian Belief Network analysis (BBN) was used to model the omnidirectional relationships of sleep duration and physical activity to history of stroke. Demographic, behavioral, health/medical, and psychosocial factors were considered as well as sleep duration [defined as short < 7hrs. and long ≥ 9hrs., referenced to healthy sleep (7-8 hrs.)], and intensity (moderate and vigorous) and frequency (times/week) of physical activity. Results: Of the sample, 48.1% were ≤45 years; 55.7% female; 77.4% were White; 15.9%, Black/African American; and 45.3% reported an annual income< $35K. Overall, the model had a precision index of 95.84%. We found that adults who reported 31-60 minutes of vigorous physical activity six times for the week and average sleep duration (7-8 hrs.) had the lowest stroke prevalence. Of the 36 sleep (short, average, and long sleep) and physical activity profiles we tested, 30 profiles had a self-reported stroke prevalence lower than the US national average of approximately 3.07%. Women, compared to men with the same sleep and physical activity profile, appeared to have higher self-reported stroke prevalence. We also report age differences across three groups 18-45, 46-65, and 66+ . Conclusion: Our findings indicate that several profiles of sleep duration and physical activity are associated with
@article{
title = {Sleep Duration and Physical Activity Profiles Associated With Self-Reported Stroke in the United States: Application of Bayesian Belief Network Modeling Techniques},
type = {article},
year = {2018},
keywords = {Stroke,gender,machine learning,physical activity,sleep duration},
pages = {534},
volume = {9},
websites = {https://www.frontiersin.org/article/10.3389/fneur.2018.00534/full},
month = {7},
publisher = {Frontiers},
day = {19},
id = {09dfc6a1-a77c-3ade-be08-a4dfd1608364},
created = {2018-07-22T20:13:09.567Z},
accessed = {2018-07-22},
file_attached = {false},
profile_id = {95e10851-cdf3-31de-9f82-1ab629e601b0},
group_id = {838ecfe2-7c01-38b2-970d-875a87910530},
last_modified = {2018-07-22T20:13:09.567Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {false},
hidden = {false},
private_publication = {false},
abstract = {Introduction: Physical activity (PA) and sleep are associated with cerebrovascular disease and events like stroke. Though the interrelationships between PA, sleep, and other stroke risk factors have been studied, we are unclear about the associations of different types, frequency and duration of PA, sleep behavioral patterns (short, average and long sleep durations), within the context of stroke-related clinical, behavioral, and socio-demographic risk factors. The current study utilized Bayesian Belief Network analysis, a type of machine learning analysis, to develop profiles of physical activity (duration, intensity, and frequency) and sleep duration associated with or no history of stroke, given the influence of multiple stroke predictors and correlates. Such a model allowed us to develop a predictive classification model of stroke which can be used in post-stroke risk stratification and developing targeted stroke rehabilitation care based on an individual’s profile. Method: Analysis was based on the 2004-2013 National Health Interview Survey (n=288,888). Bayesian Belief Network analysis (BBN) was used to model the omnidirectional relationships of sleep duration and physical activity to history of stroke. Demographic, behavioral, health/medical, and psychosocial factors were considered as well as sleep duration [defined as short < 7hrs. and long ≥ 9hrs., referenced to healthy sleep (7-8 hrs.)], and intensity (moderate and vigorous) and frequency (times/week) of physical activity. Results: Of the sample, 48.1% were ≤45 years; 55.7% female; 77.4% were White; 15.9%, Black/African American; and 45.3% reported an annual income< $35K. Overall, the model had a precision index of 95.84%. We found that adults who reported 31-60 minutes of vigorous physical activity six times for the week and average sleep duration (7-8 hrs.) had the lowest stroke prevalence. Of the 36 sleep (short, average, and long sleep) and physical activity profiles we tested, 30 profiles had a self-reported stroke prevalence lower than the US national average of approximately 3.07%. Women, compared to men with the same sleep and physical activity profile, appeared to have higher self-reported stroke prevalence. We also report age differences across three groups 18-45, 46-65, and 66+ . Conclusion: Our findings indicate that several profiles of sleep duration and physical activity are associated with},
bibtype = {article},
author = {Seixas, Azizi A. and Henclewood, Dwayne A. and Williams, Stephen K. and Jagannathan, Ram and Ramos, Alberto and Zizi, Ferdinand and Jean-Louis, Girardin},
doi = {10.3389/fneur.2018.00534},
journal = {Frontiers in Neurology}
}Super Resolution Network Analysis Defines the Molecular Architecture of Caveolae and Caveolin-1 Scaffolds. Khater, I. M., Meng, F., Wong, T. H., Nabi, I. R., & Hamarneh, G. Nature - Scientific reports, 8(9009):1-15, 2018.
doi bibtex
doi bibtex
@ARTICLE{nature_sr2018,
AUTHOR = {Ismail M. Khater and Fanrui Meng and Timothy H. Wong and
Ivan Robert Nabi and Ghassan Hamarneh},
JOURNAL = {Nature - Scientific reports},
OPTMONTH = {},
OPTNOTE = {},
NUMBER = {9009},
PAGES = {1-15},
TITLE = {Super Resolution Network Analysis Defines the Molecular
Architecture of Caveolae and Caveolin-1 Scaffolds},
VOLUME = {8},
YEAR = {2018},
OPTABSTRACT = {},
DOI = {10.1038/s41598-018-27216-4},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Super Resolution Microscopy,
Single Molecule Localization Microscopy,
Network Modelling and Analysis, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/nature_sr2018.pdf}
}seq-ImmuCC: Cell-Centric View of Tissue Transcriptome Measuring Cellular Compositions of Immune Microenvironment From Mouse RNA-Seq Data. Chen, Z., Quan, L., Huang, A., Zhao, Q., Yuan, Y., Yuan, X., Shen, Q., Shang, J., Ben, Y., Qin, F. X., & Wu, A. Frontiers in Immunology, 9:1286, 2018.
doi abstract bibtex
doi abstract bibtex
The RNA sequencing approach has been broadly used to provide gene-, pathway-, and network-centric analyses for various cell and tissue samples. However, thus far, rich cellular information carried in tissue samples has not been thoroughly characterized from RNA-Seq data. Therefore, it would expand our horizons to better understand the biological processes of the body by incorporating a cell-centric view of tissue transcriptome. Here, a computational model named seq-ImmuCC was developed to infer the relative proportions of 10 major immune cells in mouse tissues from RNA-Seq data. The performance of seq-ImmuCC was evaluated among multiple computational algorithms, transcriptional platforms, and simulated and experimental datasets. The test results showed its stable performance and superb consistency with experimental observations under different conditions. With seq-ImmuCC, we generated the comprehensive landscape of immune cell compositions in 27 normal mouse tissues and extracted the distinct signatures of immune cell proportion among various tissue types. Furthermore, we quantitatively characterized and compared 18 different types of mouse tumor tissues of distinct cell origins with their immune cell compositions, which provided a comprehensive and informative measurement for the immune microenvironment inside tumor tissues. The online server of seq-ImmuCC are freely available at http://wap-lab.org:3200/immune/.
@article{chen_seq-immucc:_2018,
title = {seq-{ImmuCC}: {Cell}-{Centric} {View} of {Tissue} {Transcriptome} {Measuring} {Cellular} {Compositions} of {Immune} {Microenvironment} {From} {Mouse} {RNA}-{Seq} {Data}},
volume = {9},
issn = {1664-3224},
shorttitle = {seq-{ImmuCC}},
doi = {10.3389/fimmu.2018.01286},
abstract = {The RNA sequencing approach has been broadly used to provide gene-, pathway-, and network-centric analyses for various cell and tissue samples. However, thus far, rich cellular information carried in tissue samples has not been thoroughly characterized from RNA-Seq data. Therefore, it would expand our horizons to better understand the biological processes of the body by incorporating a cell-centric view of tissue transcriptome. Here, a computational model named seq-ImmuCC was developed to infer the relative proportions of 10 major immune cells in mouse tissues from RNA-Seq data. The performance of seq-ImmuCC was evaluated among multiple computational algorithms, transcriptional platforms, and simulated and experimental datasets. The test results showed its stable performance and superb consistency with experimental observations under different conditions. With seq-ImmuCC, we generated the comprehensive landscape of immune cell compositions in 27 normal mouse tissues and extracted the distinct signatures of immune cell proportion among various tissue types. Furthermore, we quantitatively characterized and compared 18 different types of mouse tumor tissues of distinct cell origins with their immune cell compositions, which provided a comprehensive and informative measurement for the immune microenvironment inside tumor tissues. The online server of seq-ImmuCC are freely available at http://wap-lab.org:3200/immune/.},
language = {eng},
journal = {Frontiers in Immunology},
author = {Chen, Ziyi and Quan, Lijun and Huang, Anfei and Zhao, Qiang and Yuan, Yao and Yuan, Xuye and Shen, Qin and Shang, Jingzhe and Ben, Yinyin and Qin, F. Xiao-Feng and Wu, Aiping},
year = {2018},
pmid = {29922297},
pmcid = {PMC5996037},
keywords = {RNA-Seq, deconvolution, immune cell, machine learning, mouse, tumor},
pages = {1286},
}Natural speech algorithm applied to baseline interview data can predict which patients will respond to psilocybin for treatment-resistant depression. Carrillo, F., Sigman, M., Fernández Slezak, D., Ashton, P., Fitzgerald, L., Stroud, J., Nutt, D., J., & Carhart-Harris, R., L. Journal of Affective Disorders, 230:84-86, Elsevier, 4, 2018.
Website abstract bibtex
Website abstract bibtex Background: Natural speech analytics has seen some improvements over recent years, and this has opened a window for objective and quantitative diagnosis in psychiatry. Here, we used a machine learning algorithm applied to natural speech to ask whether language properties measured before psilocybin for treatment-resistant can predict for which patients it will be effective and for which it will not. Methods: A baseline autobiographical memory interview was conducted and transcribed. Patients with treatment-resistant depression received 2 doses of psilocybin, 10 mg and 25 mg, 7 days apart. Psychological support was provided before, during and after all dosing sessions. Quantitative speech measures were applied to the interview data from 17 patients and 18 untreated age-matched healthy control subjects. A machine learning algorithm was used to classify between controls and patients and predict treatment response. Results: Speech analytics and machine learning successfully differentiated depressed patients from healthy controls and identified treatment responders from non-responders with a significant level of 85% of accuracy (75% precision). Conclusions: Automatic natural language analysis was used to predict effective response to treatment with psilocybin, suggesting that these tools offer a highly cost-effective facility for screening individuals for treatment suitability and sensitivity. Limitations: The sample size was small and replication is required to strengthen inferences on these results.
@article{
title = {Natural speech algorithm applied to baseline interview data can predict which patients will respond to psilocybin for treatment-resistant depression},
type = {article},
year = {2018},
identifiers = {[object Object]},
keywords = {Computational psychiatry,Depression,Machine learning,Natural speech analysis,Predict therapeutic effectiveness,Psilocybin treatment,Treatment-resistant depression},
pages = {84-86},
volume = {230},
websites = {http://www.ncbi.nlm.nih.gov/pubmed/29407543,https://www.sciencedirect.com/science/article/abs/pii/S0165032717311643?via%3Dihub},
month = {4},
publisher = {Elsevier},
day = {1},
id = {7127a8f1-3589-3698-93d7-c16d49ea2bdd},
created = {2019-07-16T12:40:18.820Z},
accessed = {2018-02-13},
file_attached = {false},
profile_id = {38c6dbcb-2394-3f18-9217-58d777c08c69},
group_id = {d9389c6c-8ab5-3b8b-86ed-33db09ca0198},
last_modified = {2019-09-12T16:41:39.655Z},
tags = {CA,Disc:Psychopharmacology},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
citation_key = {Carrillo2018},
notes = {<b>From Duplicate 1 (<i>Natural speech algorithm applied to baseline interview data can predict which patients will respond to psilocybin for treatment-resistant depression</i> - Carrillo, Facundo; Sigman, Mariano; Fernández Slezak, Diego; Ashton, Philip; Fitzgerald, Lily; Stroud, Jack; Nutt, David J.; Carhart-Harris, Robin L.)<br/></b><br/>LB},
private_publication = {false},
abstract = {Background: Natural speech analytics has seen some improvements over recent years, and this has opened a window for objective and quantitative diagnosis in psychiatry. Here, we used a machine learning algorithm applied to natural speech to ask whether language properties measured before psilocybin for treatment-resistant can predict for which patients it will be effective and for which it will not. Methods: A baseline autobiographical memory interview was conducted and transcribed. Patients with treatment-resistant depression received 2 doses of psilocybin, 10 mg and 25 mg, 7 days apart. Psychological support was provided before, during and after all dosing sessions. Quantitative speech measures were applied to the interview data from 17 patients and 18 untreated age-matched healthy control subjects. A machine learning algorithm was used to classify between controls and patients and predict treatment response. Results: Speech analytics and machine learning successfully differentiated depressed patients from healthy controls and identified treatment responders from non-responders with a significant level of 85% of accuracy (75% precision). Conclusions: Automatic natural language analysis was used to predict effective response to treatment with psilocybin, suggesting that these tools offer a highly cost-effective facility for screening individuals for treatment suitability and sensitivity. Limitations: The sample size was small and replication is required to strengthen inferences on these results.},
bibtype = {article},
author = {Carrillo, Facundo and Sigman, Mariano and Fernández Slezak, Diego and Ashton, Philip and Fitzgerald, Lily and Stroud, Jack and Nutt, David J. and Carhart-Harris, Robin Lester},
journal = {Journal of Affective Disorders}
}Noninvasive Determination of Gene Mutations in Clear Cell Renal Cell Carcinoma using Multiple Instance Decisions Aggregated CNN. Hussain, A., Hamarneh, G., & Abugharbieh, R. In Lecture Notes in Computer Science, Medical Image Computing and Computer-Assisted Intervention (MICCAI), volume 11071, pages 657-665, 2018.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai2018d,
OPTADDRESS = {},
AUTHOR = {Arafat Hussain and Ghassan Hamarneh and
Rafeef Abugharbieh},
BOOKTITLE = {Lecture Notes in Computer Science, Medical Image Computing and Computer-Assisted Intervention (MICCAI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {657-665},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Noninvasive Determination of Gene Mutations in Clear
Cell Renal Cell Carcinoma using Multiple Instance Decisions
Aggregated CNN},
VOLUME = {11071},
YEAR = {2018},
OPTABSTRACT = {},
DOI = {10.1007/978-3-030-00934-2_73},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Super Resolution, Machine Learning, Deep Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai2018d.pdf}
}Generative Adversarial Networks to Segment Skin Lesions. Izadi, S., Mirikharaji, Z., Kawahara, J., & Hamarneh, G. In IEEE International Symposium on Biomedical Imaging (IEEE ISBI), pages 881-884, 2018.
doi bibtex
doi bibtex
@INPROCEEDINGS{isbi2018b,
OPTADDRESS = {},
AUTHOR = {Saeed Izadi and Zahra Mirikharaji and Jeremy Kawahara and
Ghassan Hamarneh},
BOOKTITLE = {IEEE International Symposium on Biomedical Imaging (IEEE ISBI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {881-884},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Generative Adversarial Networks to Segment Skin Lesions},
OPTVOLUME = {},
YEAR = {2018},
OPTABSTRACT = {},
DOI = {10.1109/ISBI.2018.8363712},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Dermatology, Color, Deep Learning,
Generative Adversarial Network, Segmentation},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/isbi2018b.pdf}
}Stroke-Associated Hemiparesis Detection Using Body Joints and Support Vector Machines. Ramesh, V., Agrawal, K., Meyer, B., Cauwenberghs, G., & Weibel, N. In Proceedings of the 12th EAI International Conference on Pervasive Computing Technologies for Healthcare, of PervasiveHealth '18, pages 55–58, New York, NY, USA, 2018. ACM.
Paper doi abstract bibtex
Paper doi abstract bibtex Hemiparesis, the weakness of one side of the body, affects the ability of stroke survivors to move and walk. It is generally diagnosed through motor tests performed as part of neurological examinations such as the NIH Stroke Scale (NIHSS), a subjective evaluation that requires the presence of an experienced neurologist. Here we report on an alternative way for computationally identifying hemiparesis that leverages body joint position data captured by the Microsoft Kinect. We employed support vector machines with 14 stroke subjects and 21 controls to characterize hemiparesis based on 4 core body angles recorded while the participants were simply sitting at rest, waiting for their neurologist. When comparing our results to neurologists' NIHSS scores, we were able to always identify right-side hemiparesis, left-side hemiparesis, or no hemiparesis using a leave-one-subject-out analysis. With additional data, our ultimate aim is to include the hemiparesis detection system presented here in a larger, multimodal tool that characterizes stroke based on several stroke-associated deficits. We envision deploying this tool in emergency settings for faster and more precise stroke severity assessments done in real-time.
@inproceedings{ramesh_stroke-associated_2018,
address = {New York, NY, USA},
series = {{PervasiveHealth} '18},
title = {Stroke-{Associated} {Hemiparesis} {Detection} {Using} {Body} {Joints} and {Support} {Vector} {Machines}},
isbn = {978-1-4503-6450-8},
url = {http://doi.acm.org/10.1145/3240925.3240979},
doi = {10.1145/3240925.3240979},
abstract = {Hemiparesis, the weakness of one side of the body, affects the ability of stroke survivors to move and walk. It is generally diagnosed through motor tests performed as part of neurological examinations such as the NIH Stroke Scale (NIHSS), a subjective evaluation that requires the presence of an experienced neurologist. Here we report on an alternative way for computationally identifying hemiparesis that leverages body joint position data captured by the Microsoft Kinect. We employed support vector machines with 14 stroke subjects and 21 controls to characterize hemiparesis based on 4 core body angles recorded while the participants were simply sitting at rest, waiting for their neurologist. When comparing our results to neurologists' NIHSS scores, we were able to always identify right-side hemiparesis, left-side hemiparesis, or no hemiparesis using a leave-one-subject-out analysis. With additional data, our ultimate aim is to include the hemiparesis detection system presented here in a larger, multimodal tool that characterizes stroke based on several stroke-associated deficits. We envision deploying this tool in emergency settings for faster and more precise stroke severity assessments done in real-time.},
urldate = {2018-12-06},
booktitle = {Proceedings of the 12th {EAI} {International} {Conference} on {Pervasive} {Computing} {Technologies} for {Healthcare}},
publisher = {ACM},
author = {Ramesh, Vishwajith and Agrawal, Kunal and Meyer, Brett and Cauwenberghs, Gert and Weibel, Nadir},
year = {2018},
keywords = {Body-Tracking, Hemiparesis, Kinect, Machine Learning, Posture Detection, Stroke, Support Vector Machines},
pages = {55--58},
}Star Shape Prior in Fully Convolutional Networks for Skin Lesion Segmentation. Mirikharaji, Z. & Hamarneh, G. In Lecture Notes in Computer Science, Medical Image Computing and Computer-Assisted Intervention (MICCAI), volume 11073, pages 737-745, 2018.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai2018a,
OPTADDRESS = {},
AUTHOR = {Zahra Mirikharaji and Ghassan Hamarneh},
BOOKTITLE = {Lecture Notes in Computer Science, Medical Image Computing and Computer-Assisted Intervention (MICCAI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {737-745},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Star Shape Prior in Fully Convolutional Networks for
Skin Lesion Segmentation},
VOLUME = {11073},
YEAR = {2018},
OPTABSTRACT = {},
DOI = {10.1007/978-3-030-00937-3_84},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Deep Learning, Segmentation,
Dermatology, Color, Shape Modelling and Analysis},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai2018a.pdf}
}Visually-Enabled Active Deep Learning for (Geo) Text and Image Classification: A Review. Yang, L., MacEachren, A., Mitra, P., & Onorati, T. ISPRS International Journal of Geo-Information, 7(2):65, Multidisciplinary Digital Publishing Institute, 2, 2018.
![Visually-Enabled Active Deep Learning for (Geo) Text and Image Classification: A Review [pdf]](https://bibbase.org/img/filetypes/pdf.svg "pdf")
Paper Website abstract bibtex
Paper Website abstract bibtex This paper investigates recent research on active learning for (geo) text and image classification, with an emphasis on methods that combine visual analytics and/or deep learning. Deep learning has attracted substantial attention across many domains of science and practice, because it can find intricate patterns in big data; but successful application of the methods requires a big set of labeled data. Active learning, which has the potential to address the data labeling challenge, has already had success in geospatial applications such as trajectory classification from movement data and (geo) text and image classification. This review is intended to be particularly relevant for extension of these methods to GISience, to support work in domains such as geographic information retrieval from text and image repositories, interpretation of spatial language, and related geo-semantics challenges. Specifically, to provide a structure for leveraging recent advances, we group the relevant work into five categories: active learning, visual analytics, active learning with visual analytics, active deep learning, plus GIScience and Remote Sensing (RS) using active learning and active deep learning. Each category is exemplified by recent influential work. Based on this framing and our systematic review of key research, we then discuss some of the main challenges of integrating active learning with visual analytics and deep learning, and point out research opportunities from technical and application perspectives—for application-based opportunities, with emphasis on those that address big data with geospatial components.
@article{
title = {Visually-Enabled Active Deep Learning for (Geo) Text and Image Classification: A Review},
type = {article},
year = {2018},
identifiers = {[object Object]},
keywords = {active learning,centered computing,class classification,deep learning,geographic information retrieval,human,image classification,label classification,machine learning,multi,text classification,visual analytics},
pages = {65},
volume = {7},
websites = {http://www.mdpi.com/2220-9964/7/2/65},
month = {2},
publisher = {Multidisciplinary Digital Publishing Institute},
day = {20},
id = {a6ae12c7-aee8-3ba3-850d-1bd176fa8c1c},
created = {2018-05-29T14:19:18.728Z},
accessed = {2018-05-29},
file_attached = {true},
profile_id = {6d8d7993-9618-3f6c-983a-9f6761313797},
group_id = {4f1d95d1-59ee-3ce8-85ce-055cfae2da74},
last_modified = {2018-05-29T14:19:21.156Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {false},
hidden = {false},
private_publication = {false},
abstract = {This paper investigates recent research on active learning for (geo) text and image classification, with an emphasis on methods that combine visual analytics and/or deep learning. Deep learning has attracted substantial attention across many domains of science and practice, because it can find intricate patterns in big data; but successful application of the methods requires a big set of labeled data. Active learning, which has the potential to address the data labeling challenge, has already had success in geospatial applications such as trajectory classification from movement data and (geo) text and image classification. This review is intended to be particularly relevant for extension of these methods to GISience, to support work in domains such as geographic information retrieval from text and image repositories, interpretation of spatial language, and related geo-semantics challenges. Specifically, to provide a structure for leveraging recent advances, we group the relevant work into five categories: active learning, visual analytics, active learning with visual analytics, active deep learning, plus GIScience and Remote Sensing (RS) using active learning and active deep learning. Each category is exemplified by recent influential work. Based on this framing and our systematic review of key research, we then discuss some of the main challenges of integrating active learning with visual analytics and deep learning, and point out research opportunities from technical and application perspectives—for application-based opportunities, with emphasis on those that address big data with geospatial components.},
bibtype = {article},
author = {Yang, Liping and MacEachren, Alan and Mitra, Prasenjit and Onorati, Teresa},
journal = {ISPRS International Journal of Geo-Information},
number = {2}
}Predicting Cancer with a Recurrent Visual Attention Model for Histopathology Images. BenTaieb, A. & Hamarneh, G. In Lecture Notes in Computer Science, Medical Image Computing and Computer-Assisted Intervention (MICCAI), volume 11071, pages 129-137, 2018.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai2018b,
OPTADDRESS = {},
AUTHOR = {Aicha BenTaieb and Ghassan Hamarneh},
BOOKTITLE = {Lecture Notes in Computer Science, Medical Image Computing and Computer-Assisted Intervention (MICCAI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {129-137},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Predicting Cancer with a Recurrent Visual Attention
Model for Histopathology Images},
VOLUME = {11071},
YEAR = {2018},
OPTABSTRACT = {},
DOI = {10.1007/978-3-030-00934-2_15},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Color, Machine Learning, Microscopy, Deep Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai2018b.pdf}
}Smart City Visualization Tool for the Open Data Georeferenced Analysis Utilizing Machine Learning. Estrada, E., Maciel, R., Ochoa, A., Bernabe-Loranca, B., Oliva, D., & Larios, V. International Journal of Combinatorial Optimization Problems & Informatics, 9(2):25–40, May, 2018.
Paper abstract bibtex
Paper abstract bibtex In Smart cities it is essential the development of information systems that collaborate in the measurement of the urban surroundings towards the cities' sustainability. In this research, for the key performance indicators it is proposed a pattern's visualization of efficiency metrics tool, utilizing the auto learning techniques "machine learning". The objective is to give support to the decision making throughout the georeferenced analysis exploiting the Open Data. The research was applied to the primary public schools data study case, including four stages: the study of metrics, the search of the data model, the test of territorial dependency, and the development of the tool that applies the grouping techniques or clustering to compare the development and school resources by zone. In the tool, the kmeans algorithm is implemented with label as validation method to select the more relevant centroids to display on a map.
@article{estrada_smart_2018,
title = {Smart {City} {Visualization} {Tool} for the {Open} {Data} {Georeferenced} {Analysis} {Utilizing} {Machine} {Learning}},
volume = {9},
issn = {20071558},
url = {http://ezproxy.macewan.ca/login?url=http://search.ebscohost.com/login.aspx?direct=true&db=a9h&AN=128868479&site=ehost-live&scope=site},
abstract = {In Smart cities it is essential the development of information systems that collaborate in the measurement of the urban surroundings towards the cities' sustainability. In this research, for the key performance indicators it is proposed a pattern's visualization of efficiency metrics tool, utilizing the auto learning techniques "machine learning". The objective is to give support to the decision making throughout the georeferenced analysis exploiting the Open Data. The research was applied to the primary public schools data study case, including four stages: the study of metrics, the search of the data model, the test of territorial dependency, and the development of the tool that applies the grouping techniques or clustering to compare the development and school resources by zone. In the tool, the kmeans algorithm is implemented with label as validation method to select the more relevant centroids to display on a map.},
number = {2},
urldate = {2018-09-15TZ},
journal = {International Journal of Combinatorial Optimization Problems \& Informatics},
author = {Estrada, Elsa and Maciel, Rocío and Ochoa, Alberto and Bernabe-Loranca, Beatriz and Oliva, Diego and Larios, Víctor},
month = may,
year = {2018},
keywords = {Clustering for the georeferenced analysis of the Open Data, MACHINE learning, OPEN data movement, SMART cities, Smart City Metrics for the Education Sustainability, Smart City tools},
pages = {25--40}
}Connectome Priors in Deep Neural Networks to Predict Autism (Kawahara and Brown: Joint first authors). Kawahara, J., Brown, C. J., & Hamarneh, G. In IEEE International Symposium on Biomedical Imaging (IEEE ISBI), pages 110-113, 2018.
doi bibtex
doi bibtex
@INPROCEEDINGS{isbi2018c,
OPTADDRESS = {},
AUTHOR = {Jeremy Kawahara and Colin J. Brown and Ghassan Hamarneh},
BOOKTITLE = {IEEE International Symposium on Biomedical Imaging (IEEE ISBI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {110-113},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Connectome Priors in Deep Neural Networks to Predict
Autism (Kawahara and Brown: Joint first authors)},
OPTVOLUME = {},
YEAR = {2018},
OPTABSTRACT = {},
DOI = {10.1109/ISBI.2018.8363534},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Diffusion MRI, Autism, Connectome, Machine Learning,
Deep Learning, Classification},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/isbi2018c.pdf}
}Automatic definition of robust microbiome sub-states in longitudinal data. García-Jiménez, B. & Wilkinson, M., D. PeerJ Preprints, 6:e26657v1, 3, 2018.
Website doi abstract bibtex
Website doi abstract bibtex The analysis of microbiome dynamics would allow us to elucidate patterns within microbial community evolution; however, microbiome state-transition dynamics have been scarcely studied. This is in part because a necessary first-step in such analyses has not been well-defined: how to deterministically describe a microbiome’s ”state”. Clustering in states have been widely studied, although no standard has been concluded yet. We propose a generic, domain-independent and automatic procedure to determine a reliable set of microbiome sub-states within a specific dataset, and with respect to the conditions of the study. The robustness of sub-state identification is established by the combination of diverse techniques for stable cluster verification. We reuse four distinct longitudinal microbiome datasets to demonstrate the broad applicability of our method, analysing results with different taxa subset allowing to adjust it depending on the application goal, and showing that the methodology provides a set of robust sub-states to examine in downstream studies about dynamics in microbiome.
@article{
title = {Automatic definition of robust microbiome sub-states in longitudinal data},
type = {article},
year = {2018},
keywords = {Clustering,Longitudinal dataset,Machine Learning,Metagenomics,Microbiome,Sub-states},
pages = {e26657v1},
volume = {6},
websites = {https://doi.org/10.7287/peerj.preprints.26657v1},
month = {3},
id = {8d4d3cd8-ce00-3d2b-ad84-1ea7fa6fc109},
created = {2018-10-15T08:36:22.612Z},
file_attached = {false},
profile_id = {17c87d5d-2470-32d7-b273-0734a1d9195f},
last_modified = {2018-10-15T08:36:22.612Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
citation_key = {10.7287/peerj.preprints.26657v1},
source_type = {article},
private_publication = {false},
abstract = {
The analysis of microbiome dynamics would allow us to elucidate patterns within microbial community evolution; however, microbiome state-transition dynamics have been scarcely studied. This is in part because a necessary first-step in such analyses has not been well-defined: how to deterministically describe a microbiome’s ”state”. Clustering in states have been widely studied, although no standard has been concluded yet. We propose a generic, domain-independent and automatic procedure to determine a reliable set of microbiome sub-states within a specific dataset, and with respect to the conditions of the study. The robustness of sub-state identification is established by the combination of diverse techniques for stable cluster verification. We reuse four distinct longitudinal microbiome datasets to demonstrate the broad applicability of our method, analysing results with different taxa subset allowing to adjust it depending on the application goal, and showing that the methodology provides a set of robust sub-states to examine in downstream studies about dynamics in microbiome.
},
bibtype = {article},
author = {García-Jiménez, Beatriz and Wilkinson, Mark D},
doi = {10.7287/peerj.preprints.26657v1},
journal = {PeerJ Preprints}
}Segmentation-Free Direct Tumor Volume and Metabolic Activity Estimation from PET Scans. Taghanaki, S. A., Duggan, N., Ma, H., Celler, A., Benard, F., & Hamarneh, G. Computerized Medical Imaging and Graphics (CMIG), 63(January):52-66, 2018.
doi bibtex
doi bibtex
@ARTICLE{cmig2018a,
AUTHOR = {Saeid Asgari Taghanaki and Noirin Duggan and Hillgan Ma and
Anna Celler and Francois Benard and Ghassan Hamarneh},
JOURNAL = {Computerized Medical Imaging and Graphics (CMIG)},
OPTMONTH = {},
OPTNOTE = {},
NUMBER = {January},
PAGES = {52-66},
TITLE = {Segmentation-Free Direct Tumor Volume and Metabolic
Activity Estimation from PET Scans},
VOLUME = {63},
YEAR = {2018},
OPTABSTRACT = {},
DOI = {10.1016/j.compmedimag.2017.12.004},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Segmentation,
Functional/Molecular/Dynamic Imaging},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/cmig2018a.pdf}
}Combo Loss: Handling Input and Output Imbalance in Multi-Organ Segmentation. Taghanaki, S. A., Zheng, Y., Zhou, S. K., Georgescu, B., Sharma, P., Xu, D., Comaniciu, D., & Hamarneh, G. Technical Report arxiv:1703.04559, 5, 2018.
bibtex
bibtex
@TECHREPORT{arxiv:1805.02798,
OPTADDRESS = {},
AUTHOR = {Saeid Asgari Taghanaki and Yefeng Zheng and
S. Kevin Zhou and Bogdan Georgescu and Puneet Sharma and Daguang Xu and
Dorin Comaniciu and Ghassan Hamarneh},
INSTITUTION = {},
MONTH = {5},
OPTNOTE = {},
NUMBER = {arxiv:1703.04559},
PAGES = {1-8},
TITLE = {Combo Loss: Handling Input and Output Imbalance in
Multi-Organ Segmentation},
OPTTYPE = {},
YEAR = {2018},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Deep Learning, Segmentation},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/arxiv_1805_02798.pdf}
}2017 (13)
BrainNetCNN: Artificial Convolutional Neural Networks for Connectomes. Kawahara, J., Brown, C. J., Miller, S., Booth, B. G., Chau, V., Grunau, R., Zwicker, J., & Hamarneh, G. In 2nd Annual Health Technology Symposium, Vancouver, Canada, pages 1, 2017.
bibtex
bibtex
@INPROCEEDINGS{healthtech2017c,
OPTADDRESS = {},
AUTHOR = {Jeremy Kawahara and Colin J. Brown and Steven Miller and
Brian G. Booth and Vann Chau and Ruth Grunau and Jill Zwicker and
Ghassan Hamarneh},
BOOKTITLE = {2nd Annual Health Technology Symposium, Vancouver, Canada},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {1},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {BrainNetCNN: Artificial Convolutional Neural Networks
for Connectomes},
OPTVOLUME = {},
YEAR = {2017},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Diffusion MRI, Neurodevelopment, Connectome,
Machine Learning, Deep Learning, Classification},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/healthtech2017c.pdf}
}Lesion volume Estimation from PET without Requiring Segmentation. Taghanaki, S. A., Duggan, N., Ma, H., Celler, A., Benard, F., & Hamarneh, G. In Quantitative Imaging Network (QIN) Annual Meeting, 2017.
bibtex
bibtex
@INPROCEEDINGS{qin2017,
OPTADDRESS = {},
AUTHOR = {Saeid Asgari Taghanaki and Noirin Duggan and Hillgan Ma and
Anna Celler and Francois Benard and Ghassan Hamarneh},
BOOKTITLE = {Quantitative Imaging Network (QIN) Annual Meeting},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
OPTPAGES = {},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Lesion volume Estimation from PET without Requiring
Segmentation},
OPTVOLUME = {},
YEAR = {2017},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Segmentation,
Functional/Molecular/Dynamic Imaging},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/qin2017.pdf}
}DeepCut: Object Segmentation From Bounding Box Annotations Using Convolutional Neural Networks. Rajchl, M., Lee, M. C. H., Oktay, O., Kamnitsas, K., Passerat-Palmbach, J., Bai, W., Damodaram, M., Rutherford, M. A., Hajnal, J. V., Kainz, B., & Rueckert, D. IEEE Trans. Med. Imaging, 36(2):674–683, 2017.
Paper doi bibtex
Paper doi bibtex @article{DBLP:journals/tmi/RajchlLOKPBDRHK17,
author = {Martin Rajchl and
Matthew C. H. Lee and
Ozan Oktay and
Konstantinos Kamnitsas and
Jonathan Passerat{-}Palmbach and
Wenjia Bai and
Mellisa Damodaram and
Mary A. Rutherford and
Joseph V. Hajnal and
Bernhard Kainz and
Daniel Rueckert},
title = {DeepCut: Object Segmentation From Bounding Box Annotations Using Convolutional
Neural Networks},
journal = {{IEEE} Trans. Med. Imaging},
volume = {36},
number = {2},
pages = {674--683},
year = {2017},
url = {https://doi.org/10.1109/TMI.2016.2621185},
doi = {10.1109/TMI.2016.2621185},
timestamp = {Wed, 25 Sep 2019 01:00:00 +0200},
biburl = {https://dblp.org/rec/journals/tmi/RajchlLOKPBDRHK17.bib},
bibsource = {dblp computer science bibliography, https://dblp.org}
}Learning to jump in granular media: Unifying optimal control synthesis with Gaussian process-based regression. Chang, A. H, Hubicki, C. M, Aguilar, J. J, Goldman, D. I, Ames, A. D, & Vela, P. A In Robotics and Automation (ICRA), 2017 IEEE International Conference on, pages 2154–2160, 2017. IEEE.
Paper bibtex
Paper bibtex @inproceedings{chang2017learning,
title={Learning to jump in granular media: Unifying optimal control synthesis with Gaussian process-based regression},
author={Chang, Alexander H and Hubicki, Christian M and Aguilar, Jeff J and Goldman, Daniel I and Ames, Aaron D and Vela, Patricio A},
booktitle={Robotics and Automation (ICRA), 2017 IEEE International Conference on},
pages={2154--2160},
year={2017},
organization={IEEE},
url ={http://ames.caltech.edu/chang2017learning.pdf},
keywords = {Robophysics, Machine Learning, Optimization}
}Fully Convolutional Networks to Detect Clinical Dermoscopic Features. Kawahara, J. & Hamarneh, G. Technical Report arxiv:1703.04559, 3, 2017.
bibtex
bibtex
@TECHREPORT{arxiv:1703.04559,
OPTADDRESS = {},
AUTHOR = {Jeremy Kawahara and Ghassan Hamarneh},
INSTITUTION = {},
MONTH = {3},
OPTNOTE = {},
NUMBER = {arxiv:1703.04559},
PAGES = {1-4},
TITLE = {Fully Convolutional Networks to Detect Clinical
Dermoscopic Features},
OPTTYPE = {},
YEAR = {2017},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Deep Learning, Dermatology, Challenge},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/arxiv_1703_04559.pdf}
}Exploring Stroke-associated Hemiparesis Assessment with Support Vector Machines. Ramesh, V., Agrawal, K., Meyer, B., Cauwenberghs, G., & Weibel, N. In Proceedings of the 11th EAI International Conference on Pervasive Computing Technologies for Healthcare, of PervasiveHealth '17, pages 464–467, New York, NY, USA, 2017. ACM. Poster
Paper doi abstract bibtex
Paper doi abstract bibtex Hemiparesis, the weakness of one side of the body, affects the ability of stroke survivors to move and walk. With prevalence in 80% of survivors, hemiparesis is an important measure for stroke severity. It is generally diagnosed through motor tests performed as part of the National Institute of Health Stroke Scale (NIHSS). Here we report on initial work for an alternate way of identifying hemiparesis that leverages body joint position data captured by the Microsoft Kinect v2 of people resting while waiting for the neurological examination. We employ support vector machines with 10 stroke patients and 9 healthy controls to characterize hemiparesis based on the lower core body angles of the participants, and compare our results to neurologists' diagnoses. We were able to identify left-side hemiparesis, right-side hemiparesis, or no hemiparesis with \textgreater 89% accuracy when looking at the lower body angles and observing the patients for 1 minute.
@inproceedings{ramesh_exploring_2017,
address = {New York, NY, USA},
series = {{PervasiveHealth} '17},
title = {Exploring {Stroke}-associated {Hemiparesis} {Assessment} with {Support} {Vector} {Machines}},
isbn = {978-1-4503-6363-1},
url = {http://doi.acm.org/10.1145/3154862.3154894},
doi = {10.1145/3154862.3154894},
abstract = {Hemiparesis, the weakness of one side of the body, affects the ability of stroke survivors to move and walk. With prevalence in 80\% of survivors, hemiparesis is an important measure for stroke severity. It is generally diagnosed through motor tests performed as part of the National Institute of Health Stroke Scale (NIHSS). Here we report on initial work for an alternate way of identifying hemiparesis that leverages body joint position data captured by the Microsoft Kinect v2 of people resting while waiting for the neurological examination. We employ support vector machines with 10 stroke patients and 9 healthy controls to characterize hemiparesis based on the lower core body angles of the participants, and compare our results to neurologists' diagnoses. We were able to identify left-side hemiparesis, right-side hemiparesis, or no hemiparesis with {\textgreater} 89\% accuracy when looking at the lower body angles and observing the patients for 1 minute.},
urldate = {2018-12-06},
booktitle = {Proceedings of the 11th {EAI} {International} {Conference} on {Pervasive} {Computing} {Technologies} for {Healthcare}},
publisher = {ACM},
author = {Ramesh, Vishwajith and Agrawal, Kunal and Meyer, Brett and Cauwenberghs, Gert and Weibel, Nadir},
year = {2017},
note = {Poster},
keywords = {Kinect, body-tracking, hemiparesis, machine learning, posture, stroke, support vector machines},
pages = {464--467},
}Machine-learning techniques in economics : New Tools for Predicting Economic Growth. Basuchoudhary, A., Bang, J., T., & Sen, T. Springer, Cham, 2017.
abstract bibtex
abstract bibtex
This book develops a machine-learning framework for predicting economic growth. It can also be considered as a primer for using machine learning (also known as data mining or data analytics) to answer economic questions. While machine learning itself is not a new idea, advances in computing technology combined with a dawning realization of its applicability to economic questions makes it a new tool for economists. .-- Why this Book? -- Data, Variables, and Their Sources -- Methodology -- Predicting Economic Growth: A First Look -- Predicting Economic Growth: Which Variables Matter? -- Predicting Recessions: What We Learn from Widening the Goalposts -- Epilogue.
@book{
title = {Machine-learning techniques in economics : New Tools for Predicting Economic Growth},
type = {book},
year = {2017},
identifiers = {[object Object]},
keywords = {Data mining,Econometrics,Economic growth,Forecasting,Machine learning,Prediction,Ranking predictive variables},
pages = {97},
publisher = {Springer, Cham},
city = {Cham, Switzerland},
id = {8d620585-b4c3-3412-88a8-4af1a8a1b383},
created = {2018-06-29T13:50:11.363Z},
accessed = {2018-06-29},
file_attached = {false},
profile_id = {35f49df4-e9cf-391e-a76a-8e9e316c42aa},
group_id = {ee11d323-dd56-3515-8742-9b1e2f0496f7},
last_modified = {2018-06-30T13:47:41.955Z},
read = {false},
starred = {true},
authored = {false},
confirmed = {false},
hidden = {false},
folder_uuids = {9f870d1a-3967-406b-94fb-b8524c625bf1},
private_publication = {false},
abstract = {This book develops a machine-learning framework for predicting economic growth. It can also be considered as a primer for using machine learning (also known as data mining or data analytics) to answer economic questions. While machine learning itself is not a new idea, advances in computing technology combined with a dawning realization of its applicability to economic questions makes it a new tool for economists. .-- Why this Book? -- Data, Variables, and Their Sources -- Methodology -- Predicting Economic Growth: A First Look -- Predicting Economic Growth: Which Variables Matter? -- Predicting Recessions: What We Learn from Widening the Goalposts -- Epilogue.},
bibtype = {book},
author = {Basuchoudhary, Atin and Bang, James T. and Sen, Tinni}
}A Practical Guide To Using Face Technology (Part I). Lee, F. November, 2017.
Paper abstract bibtex
Paper abstract bibtex What are the key components behind commonly used face technology?
@misc{lee_practical_2017,
title = {A {Practical} {Guide} {To} {Using} {Face} {Technology} ({Part} {I})},
url = {https://medium.com/iotforall/a-practical-guide-to-using-face-technology-part-i-7fe3fdb1bc4f},
abstract = {What are the key components behind commonly used face technology?},
urldate = {2017-11-08TZ},
journal = {Medium},
author = {Lee, Frank},
month = nov,
year = {2017},
keywords = {\#Introductory Overview, AI=Artificial Intelligence, Algorithms, Computer Science - Computer Vision and Pattern Recognition, Computer Science - Learning, Computer Vision, DNN=Deep Neural Networks, Data Science, Deep Learning, Deep Learning Frameworks, Distributed Computing, Feature Extraction, Feature Selection, Machine Learning, Machine Learning Applications, Machine Learning Frameworks, Neural Networks, Neural and Evolutionary Computing, Statistics - Machine Learning}
}Poverty from Space: Using High-Resolution Satellite Imagery for Estimating Economic Well-Being. Engstrom, R., Hersh, J., & Newhouse, D. 12 2017.
Website abstract bibtex
Website abstract bibtex Can features extracted from high spatial resolution satellite imagery accurately estimate poverty and economic well-being? This paper investigates this question by extracting object and texture features from satellite images of Sri Lanka, which are used to estimate poverty rates and average log consumption for 1,291 administrative units (Grama Niladhari divisions). The features that were extracted include the number and density of buildings, prevalence of shadows, number of cars, density and length of roads, type of agriculture, roof material, and a suite of texture and spectral features calculated using a nonoverlapping box approach. A simple linear regression model, using only these inputs as explanatory variables, explains nearly 60 percent of poverty headcount rates and average log consumption. In comparison, models built using night-time lights explain only 15 percent of the variation in poverty or income. The predictions remain accurate when restricting the sample to poorer Gram Niladhari divisions. Two sample applications, extrapolating predictions into adjacent areas and estimating local area poverty using an artificially reduced census, confirm the out-of-sample predictive capabilities.
@unpublished{
title = {Poverty from Space: Using High-Resolution Satellite Imagery for Estimating Economic Well-Being},
type = {unpublished},
year = {2017},
identifiers = {[object Object]},
keywords = {machine learning,poverty estimation,satellite imagery},
websites = {http://elibrary.worldbank.org/doi/book/10.1596/1813-9450-8284},
month = {12},
publisher = {The World Bank},
day = {19},
series = {Policy Research Working Papers},
id = {3025185a-2866-3f22-b024-973d53d56d57},
created = {2018-06-25T15:10:53.204Z},
accessed = {2018-06-25},
file_attached = {false},
profile_id = {35f49df4-e9cf-391e-a76a-8e9e316c42aa},
group_id = {ee11d323-dd56-3515-8742-9b1e2f0496f7},
last_modified = {2018-07-02T08:33:32.243Z},
read = {false},
starred = {true},
authored = {false},
confirmed = {false},
hidden = {false},
folder_uuids = {a77eb13c-f2a5-421f-855f-30c90585d1bc,0f486643-3d08-4632-8e8e-e53433d903a9},
private_publication = {false},
abstract = {Can features extracted from high spatial resolution satellite imagery accurately estimate poverty and economic well-being? This paper investigates this question by extracting object and texture features from satellite images of Sri Lanka, which are used to estimate poverty rates and average log consumption for 1,291 administrative units (Grama Niladhari divisions). The features that were extracted include the number and density of buildings, prevalence of shadows, number of cars, density and length of roads, type of agriculture, roof material, and a suite of texture and spectral features calculated using a nonoverlapping box approach. A simple linear regression model, using only these inputs as explanatory variables, explains nearly 60 percent of poverty headcount rates and average log consumption. In comparison, models built using night-time lights explain only 15 percent of the variation in poverty or income. The predictions remain accurate when restricting the sample to poorer Gram Niladhari divisions. Two sample applications, extrapolating predictions into adjacent areas and estimating local area poverty using an artificially reduced census, confirm the out-of-sample predictive capabilities.},
bibtype = {unpublished},
author = {Engstrom, Ryan and Hersh, Jonathan and Newhouse, David}
}Molecular Level Quantification of Cav1 Clusters in Super-Resolution Imaging Data. Khater, I. M., Meng, F., Nabi, I. R., & Hamarneh, G. In Frontiers in Biophysics, Vancouver, Canada, pages 1, 2017.
bibtex
bibtex
@INPROCEEDINGS{frontiers_biophysics2017,
OPTADDRESS = {},
AUTHOR = {Ismail M. Khater and Fanrui Meng and Ivan Robert Nabi and
Ghassan Hamarneh},
BOOKTITLE = {Frontiers in Biophysics, Vancouver, Canada},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {1},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Molecular Level Quantification of Cav1 Clusters in
Super-Resolution Imaging Data},
OPTVOLUME = {},
YEAR = {2017},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Super Resolution Microscopy,
Single Molecule Localization Microscopy,
Network Modelling and Analysis, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/frontiers_biophysics2017.pdf}
}Counting Apples and Oranges With Deep Learning: A Data-Driven Approach. Chen, S. W., Shivakumar, S. S., Dcunha, S., Das, J., Okon, E., Qu, C., Taylor, C. J., & Kumar, V. IEEE Robotics and Automation Letters, 2(2):781–788, April, 2017.
doi abstract bibtex
doi abstract bibtex
This paper describes a fruit counting pipeline based on deep learning that accurately counts fruit in unstructured environments. Obtaining reliable fruit counts is challenging because of variations in appearance due to illumination changes and occlusions from foliage and neighboring fruits. We propose a novel approach that uses deep learning to map from input images to total fruit counts. The pipeline utilizes a custom crowdsourcing platform to quickly label large data sets. A blob detector based on a fully convolutional network extracts candidate regions in the images. A counting algorithm based on a second convolutional network then estimates the number of fruits in each region. Finally, a linear regression model maps that fruit count estimate to a final fruit count. We analyze the performance of the pipeline on two distinct data sets of oranges in daylight, and green apples at night, utilizing human generated labels as ground truth. We also show that the pipeline has a short training time and performs well with a limited data set size. Our method generalizes across both data sets and is able to perform well even on highly occluded fruits that are challenging for human labelers to annotate.
@article{chen_counting_2017,
title = {Counting {Apples} and {Oranges} {With} {Deep} {Learning}: {A} {Data}-{Driven} {Approach}},
volume = {2},
shorttitle = {Counting {Apples} and {Oranges} {With} {Deep} {Learning}},
doi = {10.1109/LRA.2017.2651944},
abstract = {This paper describes a fruit counting pipeline based on deep learning that accurately counts fruit in unstructured environments. Obtaining reliable fruit counts is challenging because of variations in appearance due to illumination changes and occlusions from foliage and neighboring fruits. We propose a novel approach that uses deep learning to map from input images to total fruit counts. The pipeline utilizes a custom crowdsourcing platform to quickly label large data sets. A blob detector based on a fully convolutional network extracts candidate regions in the images. A counting algorithm based on a second convolutional network then estimates the number of fruits in each region. Finally, a linear regression model maps that fruit count estimate to a final fruit count. We analyze the performance of the pipeline on two distinct data sets of oranges in daylight, and green apples at night, utilizing human generated labels as ground truth. We also show that the pipeline has a short training time and performs well with a limited data set size. Our method generalizes across both data sets and is able to perform well even on highly occluded fruits that are challenging for human labelers to annotate.},
number = {2},
journal = {IEEE Robotics and Automation Letters},
author = {Chen, S. W. and Shivakumar, S. S. and Dcunha, S. and Das, J. and Okon, E. and Qu, C. and Taylor, C. J. and Kumar, V.},
month = apr,
year = {2017},
keywords = {Agricultural automation, Image segmentation, Labeling, Lighting, Machine learning, Neural networks, Pipelines, Training, agricultural automation, agricultural products, apple counting, blob detector, candidate region extraction, categorization, counting algorithm, crowdsourcing, crowdsourcing platform, data-driven approach, feature extraction, feedforward neural nets, fruit counting pipeline, fully convolutional network, green apples, human generated label utilization, illumination changes, large data sets, learning (artificial intelligence), linear regression model, object detection, occlusions, orange counting, performance analysis, performance evaluation, regression analysis, segmentation, visual learning},
pages = {781--788}
}Uncertainty Driven Multi-Loss Fully Convolutional Networks for Gland Analysis. BenTaieb, A. & Hamarneh, G. In Medical Image Computing and Computer-Assisted Intervention Workshop on Large-scale Annotation of Biomedical data and Expert Label Synthesis (MICCAI LABELS), volume 10552, pages 155-163, 2017.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai_labels2017a,
OPTADDRESS = {},
AUTHOR = {Aicha BenTaieb and Ghassan Hamarneh},
BOOKTITLE = {Medical Image Computing and Computer-Assisted Intervention Workshop on Large-scale Annotation of Biomedical data and Expert Label Synthesis (MICCAI LABELS)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {155-163},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Uncertainty Driven Multi-Loss Fully Convolutional
Networks for Gland Analysis},
VOLUME = {10552},
YEAR = {2017},
OPTABSTRACT = {},
DOI = {10.1007/978-3-319-67534-3_17},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Color, Machine Learning, Microscopy, Deep Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai_labels2017a.pdf}
}A Toolkit for Ecosystem Ecologists in the Time of Big Science. Peters, D. P. & Okin, G. S. Ecosystems, 20(2):259–266, March, 2017.
Paper doi abstract bibtex
Paper doi abstract bibtex Ecosystem ecologists are being challenged to address the increasingly complex problems that comprise Big Science. These problems include multiple levels of biological organization that cross multiple interacting temporal and spatial scales, from individual plants, animals, and microbes to landscapes, continents, and the globe. As technology improves, the availability of data, derived data products, and information to address these complex problems are increasing at finer and coarser scales of resolution, and legacy, dark data are brought to light. Data analytics are improving as big data increase in importance in other fields that are improving access to these data. New data sources (crowdsourcing, social media) and ease of communication and collaboration among ecosystem ecologists and other disciplines are increasingly possible via the internet. It is increasingly important that ecosystem ecologists be able to communicate their findings, and to translate their concepts and findings into concrete bits of information that a general public can understand. Traditional approaches that portray ecosystem sciences as a dichotomy between empirical research and theoretical research will keep the field from fully contributing to the complexity of global change questions, and will keep ecosystem ecologists from taking full advantage of the data and technology available. Building on previous research, we describe a more forward-looking, integrated empirical–theoretical modeling approach that is iterative with learning to take advantage of the elements of Big Science. We suggest that training ecosystem ecologists in this integrated approach will be critical to addressing complex Earth system science questions, now and in the future.
@article{peters_toolkit_2017,
title = {A {Toolkit} for {Ecosystem} {Ecologists} in the {Time} of {Big} {Science}},
volume = {20},
issn = {1435-0629},
url = {https://doi.org/10.1007/s10021-016-0072-1},
doi = {10.1007/s10021-016-0072-1},
abstract = {Ecosystem ecologists are being challenged to address the increasingly complex problems that comprise Big Science. These problems include multiple levels of biological organization that cross multiple interacting temporal and spatial scales, from individual plants, animals, and microbes to landscapes, continents, and the globe. As technology improves, the availability of data, derived data products, and information to address these complex problems are increasing at finer and coarser scales of resolution, and legacy, dark data are brought to light. Data analytics are improving as big data increase in importance in other fields that are improving access to these data. New data sources (crowdsourcing, social media) and ease of communication and collaboration among ecosystem ecologists and other disciplines are increasingly possible via the internet. It is increasingly important that ecosystem ecologists be able to communicate their findings, and to translate their concepts and findings into concrete bits of information that a general public can understand. Traditional approaches that portray ecosystem sciences as a dichotomy between empirical research and theoretical research will keep the field from fully contributing to the complexity of global change questions, and will keep ecosystem ecologists from taking full advantage of the data and technology available. Building on previous research, we describe a more forward-looking, integrated empirical–theoretical modeling approach that is iterative with learning to take advantage of the elements of Big Science. We suggest that training ecosystem ecologists in this integrated approach will be critical to addressing complex Earth system science questions, now and in the future.},
language = {en},
number = {2},
urldate = {2019-02-25},
journal = {Ecosystems},
author = {Peters, Debra P.C. and Okin, Gregory S.},
month = mar,
year = {2017},
keywords = {LTER, LTER-JRN, analytics, article, big data, crowdsourcing, interacting spatial and temporal scales, journal, machine learning, multiple levels of organization, technological Advances, technological advances},
pages = {259--266}
}2016 (16)
S/HIC: Robust Identification of Soft and Hard Sweeps Using Machine Learning. Schrider, D. R. & Kern, A. D. PLOS Genetics, 12(3):e1005928, March, 2016.
Paper doi abstract bibtex
Paper doi abstract bibtex Detecting the targets of adaptive natural selection from whole genome sequencing data is a central problem for population genetics. However, to date most methods have shown sub-optimal performance under realistic demographic scenarios. Moreover, over the past decade there has been a renewed interest in determining the importance of selection from standing variation in adaptation of natural populations, yet very few methods for inferring this model of adaptation at the genome scale have been introduced. Here we introduce a new method, S/HIC, which uses supervised machine learning to precisely infer the location of both hard and soft selective sweeps. We show that S/HIC has unrivaled accuracy for detecting sweeps under demographic histories that are relevant to human populations, and distinguishing sweeps from linked as well as neutrally evolving regions. Moreover, we show that S/HIC is uniquely robust among its competitors to model misspecification. Thus, even if the true demographic model of a population differs catastrophically from that specified by the user, S/HIC still retains impressive discriminatory power. Finally, we apply S/HIC to the case of resequencing data from human chromosome 18 in a European population sample, and demonstrate that we can reliably recover selective sweeps that have been identified earlier using less specific and sensitive methods.
@article{schrider_s/hic:_2016,
title = {S/{HIC}: {Robust} {Identification} of {Soft} and {Hard} {Sweeps} {Using} {Machine} {Learning}},
volume = {12},
issn = {1553-7404},
shorttitle = {S/{HIC}},
url = {https://journals.plos.org/plosgenetics/article?id=10.1371/journal.pgen.1005928},
doi = {10.1371/journal.pgen.1005928},
abstract = {Detecting the targets of adaptive natural selection from whole genome sequencing data is a central problem for population genetics. However, to date most methods have shown sub-optimal performance under realistic demographic scenarios. Moreover, over the past decade there has been a renewed interest in determining the importance of selection from standing variation in adaptation of natural populations, yet very few methods for inferring this model of adaptation at the genome scale have been introduced. Here we introduce a new method, S/HIC, which uses supervised machine learning to precisely infer the location of both hard and soft selective sweeps. We show that S/HIC has unrivaled accuracy for detecting sweeps under demographic histories that are relevant to human populations, and distinguishing sweeps from linked as well as neutrally evolving regions. Moreover, we show that S/HIC is uniquely robust among its competitors to model misspecification. Thus, even if the true demographic model of a population differs catastrophically from that specified by the user, S/HIC still retains impressive discriminatory power. Finally, we apply S/HIC to the case of resequencing data from human chromosome 18 in a European population sample, and demonstrate that we can reliably recover selective sweeps that have been identified earlier using less specific and sensitive methods.},
language = {en},
number = {3},
urldate = {2019-03-01TZ},
journal = {PLOS Genetics},
author = {Schrider, Daniel R. and Kern, Andrew D.},
month = mar,
year = {2016},
keywords = {Decision trees, Europe, Genomics statistics, Haplotypes, Machine learning, Population genetics, Population size, Simulation and modeling},
pages = {e1005928}
}Machine Learning is Fun! Part 3: Deep Learning and Convolutional Neural Networks. Geitgey, A. June, 2016.
Paper abstract bibtex
Paper abstract bibtex Update: This article is part of a series. Check out the full series: Part 1, Part 2, Part 3, Part 4, Part 5, Part 6, Part 7 and Part 8!
@misc{geitgey_machine_2016-1,
title = {Machine {Learning} is {Fun}! {Part} 3: {Deep} {Learning} and {Convolutional} {Neural} {Networks}},
shorttitle = {Machine {Learning} is {Fun}! {Part} 3},
url = {https://medium.com/@ageitgey/machine-learning-is-fun-part-3-deep-learning-and-convolutional-neural-networks-f40359318721},
abstract = {Update: This article is part of a series. Check out the full series: Part 1, Part 2, Part 3, Part 4, Part 5, Part 6, Part 7 and Part 8!},
urldate = {2018-05-03TZ},
journal = {Medium},
author = {Geitgey, Adam},
month = jun,
year = {2016},
keywords = {\#Introductory Overview, \#KeyReference, AI=Artificial Intelligence, Algorithms, CNN=Convolutional Neural Network, Classification, Computer Science, Computer Science - Computer Vision and Pattern Recognition, Computer Vision, DNN=Deep Neural Networks, Data Science, Deep Learning, GPU Computing, Image Analysis, Informatics, Machine Learning, Machine Learning Applications, Machine Learning Frameworks, Max Pooling Algorithm, Neural Networks, Python (Computer Programming), Sliding Input Window Technique, Source Code, Statistics - Machine Learning, TensorFlow, Working Code}
}Discovering Biosignatures of Cav1 Domains: Computational Methods for Super-resolution Microscopy. Khater, I. M., Meng, F., Nabi, I. R., & Hamarneh, G. In LSI Imaging Super-resolution Mini-symposium, Vancouver, Canada, pages 1, 2016.
bibtex
bibtex
@INPROCEEDINGS{superres2016,
OPTADDRESS = {},
AUTHOR = {Ismail M. Khater and Fanrui Meng and Ivan Robert Nabi and
Ghassan Hamarneh},
BOOKTITLE = {LSI Imaging Super-resolution Mini-symposium, Vancouver, Canada},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {1},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Discovering Biosignatures of Cav1 Domains: Computational
Methods for Super-resolution Microscopy},
OPTVOLUME = {},
YEAR = {2016},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Super Resolution Microscopy,
Single Molecule Localization Microscopy,
Network Modelling and Analysis, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/superres2016.pdf}
}Clinically-Inspired Automatic Classification of Ovarian Carcinoma Subtypes. BenTaieb, A., Nosrati, M., Li-Chang, H., Huntsman, D., & Hamarneh, G. Journal of Pathology Informatics, 7(1):1-28, 2016.
doi bibtex
doi bibtex
@ARTICLE{jpi2016,
AUTHOR = {Aicha BenTaieb and Masoud Nosrati and Hector Li-Chang and
David Huntsman and Ghassan Hamarneh},
JOURNAL = {Journal of Pathology Informatics},
OPTMONTH = {},
OPTNOTE = {},
NUMBER = {1},
PAGES = {1-28},
TITLE = {Clinically-Inspired Automatic Classification of Ovarian
Carcinoma Subtypes},
VOLUME = {7},
YEAR = {2016},
OPTABSTRACT = {},
DOI = {10.4103/2153-3539.186899},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Color, Machine Learning, Microscopy},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/jpi2016.pdf}
}Automated detection and classification of feeding strikes by larval fish from continuous high-speed digital video: a novel method to extract quantitative data from fast, sparse kinematic events. Zilka, M., Eyal Shamur, E., Hassner, T., China, V., Liberzon, A., & Holzman, R. Journal of Experimental Biology, 2016.
Paper abstract bibtex
Paper abstract bibtex Using videography to extract quantitative data on animal movement and kinematics constitutes a major tool in biomechanics and behavioral ecology. Advanced recording technologies now enable acquisition of long video sequences encompassing sparse and unpredictable events. Although such events may be ecologically important, analysis of sparse data can be extremely time-consuming and potentially biased; data quality is often strongly dependent on the training level of the observer and subject to contamination by observer-dependent biases. These constraints often limit our ability to study animal performance and fitness. Using long videos of foraging fish larvae, we provide a framework for the automated detection of prey acquisition strikes, a behavior that is infrequent yet critical for larval survival.We compared the performance of four video descriptors and their combinations against manually identified feeding events. For our data, the best single descriptor provided a classification accuracy of 77-95% and detection accuracy of 88- 98%, depending on fish species and size. Using a combination of descriptors improved the accuracy of classification by ∼2%, but did not improve detection accuracy. Our results indicate that the effort required by an expert to manually label videos can be greatly reduced to examining only the potential feeding detections in order to filter false detections. Thus, using automated descriptors reduces the amount of manual work needed to identify events of interest from weeks to hours, enabling the assembly of an unbiased large dataset of ecologically relevant behaviors.
@article{
title = {Automated detection and classification of feeding strikes by larval fish from continuous high-speed digital video: a novel method to extract quantitative data from fast, sparse kinematic events},
type = {article},
year = {2016},
identifiers = {[object Object]},
keywords = {automated classification,feeding kinematics,high-speed video,machine learning},
pages = {1608-1617},
id = {8cd4dbcd-9f4b-351d-a3fe-175da1b03bb4},
created = {2016-09-28T17:56:50.000Z},
accessed = {2016-09-28},
file_attached = {true},
profile_id = {fb329fcb-394b-3686-a7d8-4373fd71ca2d},
last_modified = {2017-03-13T16:27:28.843Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
citation_key = {Shamur2016},
private_publication = {false},
abstract = {Using videography to extract quantitative data on animal movement and kinematics constitutes a major tool in biomechanics and behavioral ecology. Advanced recording technologies now enable acquisition of long video sequences encompassing sparse and unpredictable events. Although such events may be ecologically important, analysis of sparse data can be extremely time-consuming and potentially biased; data quality is often strongly dependent on the training level of the observer and subject to contamination by observer-dependent biases. These constraints often limit our ability to study animal performance and fitness. Using long videos of foraging fish larvae, we provide a framework for the automated detection of prey acquisition strikes, a behavior that is infrequent yet critical for larval survival.We compared the performance of four video descriptors and their combinations against manually identified feeding events. For our data, the best single descriptor provided a classification accuracy of 77-95% and detection accuracy of 88- 98%, depending on fish species and size. Using a combination of descriptors improved the accuracy of classification by ∼2%, but did not improve detection accuracy. Our results indicate that the effort required by an expert to manually label videos can be greatly reduced to examining only the potential feeding detections in order to filter false detections. Thus, using automated descriptors reduces the amount of manual work needed to identify events of interest from weeks to hours, enabling the assembly of an unbiased large dataset of ecologically relevant behaviors.},
bibtype = {article},
author = {Zilka, Miri and Eyal Shamur, E and Hassner, Tal and China, Victor and Liberzon, Alex and Holzman, Roi},
journal = {Journal of Experimental Biology}
}Backcasting and a new way of command in computational design. Koenig, R. & Schmitt, G. In CAADence in Architecture, pages 15–25, 2016.
doi abstract bibtex
doi abstract bibtex
It's not uncommon that analysis and simulation methods are used mainly to evaluate finished designs and to proof their quality. Whereas the potential of such methods is to lead or control a design process from the beginning on. Therefore, we introduce a design method that move away from a “what-if” forecasting philosophy and increase the focus on backcasting approaches. We use the power of computation by combining sophisticated methods to generate design with analysis methods to close the gap between analysis and synthesis of designs. For the development of a future-oriented computational design support we need to be aware of the human designer’s role. A productive combination of the excellence of human cognition with the power of modern computing technology is needed. We call this approach “cognitive design computing”. The computational part aim to mimic the way a designer’s brain works by combining state-of-the-art optimization and machine learning approaches with available simulation methods. The cognition part respects the complex nature of design problems by the provision of models for human-computation interaction. This means that a design problem is distributed between computer and designer. In the context of the conference slogan “back to command”, we ask how we may imagine the command over a cognitive design computing system. We expect that designers will need to let go control of some parts of the design process to machines, but in exchange they will get a new powerful command on complex computing processes. This means that designers have to explore the potentials of their role as commanders of partially automated design processes. In this contribution we describe an approach for the development of a future cognitive design computing system with the focus on urban design issues. The aim of this system is to enable an urban planner to treat a planning problem as a backcasting problem by defining what performance a design solution should achieve and to automatically query or generate a set of best possible solutions. This kind of computational planning process offers proof that the designer meets the original explicitly defined design requirements. A key way in which digital tools can support designers is by generating design proposals. Evolutionary multi-criteria optimization methods allow us to explore a multi-dimensional design space and provide a basis for the designer to evaluate contradicting requirements: a task urban planners are faced with frequently. We also reflect why designers will give more and more control to machines. Therefore, we investigate first approaches learn how designers use computational design support systems in combination with manual design strategies to deal with urban design problems by employing machine learning methods. By observing how designers work, it is possible to derive more complex artificial solution strategies that can help computers make better suggestions in the future.
@inproceedings{koenig_backcasting_2016,
title = {Backcasting and a new way of command in computational design},
copyright = {All rights reserved},
doi = {10.3311/CAADence.1692},
abstract = {It's not uncommon that analysis and simulation methods are used mainly to evaluate finished designs and to proof their quality. Whereas the potential of such methods is to lead or control a design process from the beginning on. Therefore, we introduce a design method that move away from a “what-if” forecasting philosophy and increase the focus on backcasting approaches. We use the power of computation by combining sophisticated methods to generate design with analysis methods to close the gap between analysis and synthesis of designs. For the development of a future-oriented computational design support we need to be aware of the human designer’s role. A productive combination of the excellence of human cognition with the power of modern computing technology is needed. We call this approach “cognitive design computing”. The computational part aim to mimic the way a designer’s brain works by combining state-of-the-art optimization and machine learning approaches with available simulation methods. The cognition part respects the complex nature of design problems by the provision of models for human-computation interaction. This means that a design problem is distributed between computer and designer. In the context of the conference slogan “back to command”, we ask how we may imagine the command over a cognitive design computing system. We expect that designers will need to let go control of some parts of the design process to machines, but in exchange they will get a new powerful command on complex computing processes. This means that designers have to explore the potentials of their role as commanders of partially automated design processes. In this contribution we describe an approach for the development of a future cognitive design computing system with the focus on urban design issues. The aim of this system is to enable an urban planner to treat a planning problem as a backcasting problem by defining what performance a design solution should achieve and to automatically query or generate a set of best possible solutions. This kind of computational planning process offers proof that the designer meets the original explicitly defined design requirements. A key way in which digital tools can support designers is by generating design proposals. Evolutionary multi-criteria optimization methods allow us to explore a multi-dimensional design space and provide a basis for the designer to evaluate contradicting requirements: a task urban planners are faced with frequently. We also reflect why designers will give more and more control to machines. Therefore, we investigate first approaches learn how designers use computational design support systems in combination with manual design strategies to deal with urban design problems by employing machine learning methods. By observing how designers work, it is possible to derive more complex artificial solution strategies that can help computers make better suggestions in the future.},
booktitle = {{CAADence} in {Architecture}},
author = {Koenig, Reinhard and Schmitt, Gerhard},
year = {2016},
keywords = {Cognitive design computing, backcasting, design synthesis, evolutionary optimization, machine learning},
pages = {15--25},
}Tumour Lesion Segmentation from 3D PET using a Machine Learning driven Active Surface. Ahmadvand, P., Duggan, N., Benard, F., & Hamarneh, G. In Medical Image Computing and Computer-Assisted Intervention Workshop on Machine Learning in Medical Imaging (MICCAI MLMI), volume 10019, pages 271-278, 2016.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai_mlmi2016b,
OPTADDRESS = {},
AUTHOR = {Payam Ahmadvand and Noirin Duggan and Francois Benard and
Ghassan Hamarneh},
BOOKTITLE = {Medical Image Computing and Computer-Assisted Intervention Workshop on Machine Learning in Medical Imaging (MICCAI MLMI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {271-278},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Tumour Lesion Segmentation from 3D PET using a Machine
Learning driven Active Surface},
VOLUME = {10019},
YEAR = {2016},
OPTABSTRACT = {},
DOI = {10.1007/978-3-319-47157-0_33},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Segmentation,
Functional/Molecular/Dynamic Imaging},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai_mlmi2016b.pdf}
}A Data-Driven Demand Response Recommender System. Behl, M. & Mangharam, R. Journal of Applied Energy, 2016. [Under Review]
abstract bibtex
abstract bibtex
Demand response (DR) is becoming increasingly important as the volatility on the grid continues to increase. Current DR approaches are predominantly completely manual and rule-based or involve deriving first principles based models which are extremely cost and time prohibitive to build. We consider the problem of data-driven end-user DR for large buildings which involves predicting the demand response baseline, evaluating fixed rule based DR strategies and synthesizing DR control actions. The challenge is in evaluating and taking control decisions at fast time scales in order to curtail the power consumption of the building, in return for a financial reward. We provide a model based control with regression trees algorithm (mbCRT), which allows us to perform closed-loop control for DR strategy synthesis for large commercial buildings. Our data-driven control synthesis algorithm outperforms rule-based DR by 17% for a large DoE commercial reference building and leads to a curtailment of 380kW and over $45, 000 in savings. Our methods have been integrated into an open source tool called DR-Advisor, which acts as a recommender system for the building’s facilities manager and provides suitable control actions to meet the desired load curtailment while main- taining operations and maximizing the economic reward. DR-Advisor achieves 92.8% to 98.9% prediction accuracy for 8 buildings on Penn’s campus. We compare DR-Advisor with other data driven methods and rank 2nd on ASHRAE’s benchmarking data-set for energy prediction.
@Article{BehlMangharam2015,
author = {Madhur Behl and Rahul Mangharam},
title = {A Data-Driven Demand Response Recommender System},
journal = {Journal of Applied Energy},
year = {2016},
note = {[Under Review]},
__markedentry = {[Madhur:6]},
abstract = {Demand response (DR) is becoming increasingly important as the volatility on the grid continues to increase. Current DR approaches are predominantly completely manual and rule-based or involve deriving first principles based models which are extremely cost and time prohibitive to build. We consider the problem of data-driven end-user DR for large buildings which involves predicting the demand response baseline, evaluating fixed rule based DR strategies and synthesizing DR control actions. The challenge is in evaluating and taking control decisions at fast time scales in order to curtail the power consumption of the building, in return for a financial reward. We provide a model based control with regression trees algorithm (mbCRT), which allows us to perform closed-loop control for DR strategy synthesis for large commercial buildings. Our data-driven control synthesis algorithm outperforms rule-based DR by 17% for a large DoE commercial reference building and leads to a curtailment of 380kW and over $45, 000 in savings. Our methods have been integrated into an open source tool called DR-Advisor, which acts as a recommender system for the building’s facilities manager and provides suitable control actions to meet the desired load curtailment while main- taining operations and maximizing the economic reward. DR-Advisor achieves 92.8% to 98.9% prediction accuracy for 8 buildings on Penn’s campus. We compare DR-Advisor with other data driven methods and rank 2nd on ASHRAE’s benchmarking data-set for energy prediction.},
keywords = {Demand Response, Regression trees, data-driven control, machine learning, Electricity curtailment, Demand side management}
}Predictive Subnetwork Extraction with Structural Priors for Infant Connectomes. Brown, C. J., Miller, S., Booth, B. G., Zwicker, J., Grunau, R., Synnes, A., Chau, V., & Hamarneh, G. In Lecture Notes in Computer Science, Medical Image Computing and Computer-Assisted Intervention (MICCAI), volume 9900, pages 175-183, 2016.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai2016b,
OPTADDRESS = {},
AUTHOR = {Colin J. Brown and Steven Miller and Brian G. Booth and
Jill Zwicker and Ruth Grunau and Anne Synnes and Vann Chau and
Ghassan Hamarneh},
BOOKTITLE = {Lecture Notes in Computer Science, Medical Image Computing and Computer-Assisted Intervention (MICCAI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {175-183},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Predictive Subnetwork Extraction with Structural Priors
for Infant Connectomes},
VOLUME = {9900},
YEAR = {2016},
OPTABSTRACT = {},
DOI = {10.1007/978-3-319-46720-7_21},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Neurodevelopment, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai2016b.pdf}
}Learning Time-Varying Forecast Combinations. Mandel, A. & Sani, A. Working Paper Centre d'Economie de la Sorbonne 2016.36, 2016.
Paper abstract bibtex
Paper abstract bibtex Non-parametric forecast combination methods choose a set of static weights to combine over candidate forecasts as opposed to traditional forecasting approaches, such as ordinary least squares, that combine over information (e.g. exogenous variables). While they are robust to noise, structural breaks, inconsistent predictors and changing dynamics in the target variable, sophisticated combination methods fail to outperform the simple mean. Time-varying weights have been suggested as a way forward. Here we address the challenge to “develop methods better geared to the intermittent and evolving nature of predictive relations” in Stock and Watson (2001) and propose a data driven machine learning approach to learn time-varying forecast combinations for output, inflation or any macroeconomic time series of interest. Further, the proposed procedure “hedges” combination weights against poor performance to the mean, while optimizing weights to minimize the performance gap to the best candidate forecast in hindsight. Theoretical results are reported along with empirical performance on a standard macroeconomic dataset for predicting output and inflation.
@article{Mandel2016b,
abstract = {Non-parametric forecast combination methods choose a set of static weights to combine over candidate forecasts as opposed to traditional forecasting approaches, such as ordinary least squares, that combine over information (e.g. exogenous variables). While they are robust to noise, structural breaks, inconsistent predictors and changing dynamics in the target variable, sophisticated combination methods fail to outperform the simple mean. Time-varying weights have been suggested as a way forward. Here we address the challenge to “develop methods better geared to the intermittent and evolving nature of predictive relations” in Stock and Watson (2001) and propose a data driven machine learning approach to learn time-varying forecast combinations for output, inflation or any macroeconomic time series of interest. Further, the proposed procedure “hedges” combination weights against poor performance to the mean, while optimizing weights to minimize the performance gap to the best candidate forecast in hindsight. Theoretical results are reported along with empirical performance on a standard macroeconomic dataset for predicting output and inflation.},
author = {Mandel, Antoine and Sani, Amir},
journal = {Working Paper Centre d'Economie de la Sorbonne 2016.36},
keywords = {DOLFINS{\_}T1.3,DOLFINS{\_}T2.2,DOLFINS{\_}T2.3,DOLFINS{\_}WP1,DOLFINS{\_}WP2,DOLFINS{\_}working{\_}paper},
mendeley-tags = {DOLFINS{\_}T1.3,DOLFINS{\_}T2.2,DOLFINS{\_}T2.3,DOLFINS{\_}WP1,DOLFINS{\_}WP2,DOLFINS{\_}working{\_}paper},
title = {{Learning Time-Varying Forecast Combinations}},
url = {https://halshs.archives-ouvertes.fr/halshs-01317974/},
year = {2016}
}Comparing of feature selection and classification methods on report-based subhealth data. Li Huang, Shixing Yan, Jiamin Yuan, Zhiya Zuo, Fuping Xu, Yanzhao Lin, Mary Qu Yang, Zhimin Yang, & Li, G. In 2016 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), pages 1356-1358, 12, 2016. IEEE.
Website abstract bibtex
Website abstract bibtex © 2016 IEEE.Sub-health is a state between health and disease conditions, which is common among people living with the fierce competition and rapid pace of modern life. At present, there are no unified approaches to diagnose the sub-health patients. Self-reporting, the use of questionnaires, is one of the most popular approaches to evaluate health conditions. While a questionnaire consists of as many as 400 questions, people are likely to lose patience. This paper presents a machine learning method to mine the sub-health related questions and then provide classification suggestion based on the self-reporting data collected from Sub-health Condition Identification and Classification Research project. To study the most effective mining approaches, four different feature selection methods were applied to discovery the internal relationship among questions and four different supervised learning classifiers were utilized to investigate the most related questions to the specific diagnostic tasks. Experimental results show that artificial neural network achieves the best performance and the final diagnostic accuracy reaches 84.07% with 20 most related questions.
@inProceedings{
title = {Comparing of feature selection and classification methods on report-based subhealth data},
type = {inProceedings},
year = {2016},
identifiers = {[object Object]},
keywords = {Classification,Feature selection,Machine learning,Self-reporting,Sub-health},
pages = {1356-1358},
websites = {http://ieeexplore.ieee.org/document/7822716/},
month = {12},
publisher = {IEEE},
institution = {IEEE},
id = {62aea12b-7899-3586-a8c2-98e9168591ca},
created = {2017-10-24T02:47:20.615Z},
file_attached = {false},
profile_id = {f4ee1785-e903-38e8-befc-faa9e9b331d1},
last_modified = {2018-02-26T18:48:33.727Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
citation_key = {huang2016comparing},
source_type = {inproceedings},
private_publication = {false},
abstract = {© 2016 IEEE.Sub-health is a state between health and disease conditions, which is common among people living with the fierce competition and rapid pace of modern life. At present, there are no unified approaches to diagnose the sub-health patients. Self-reporting, the use of questionnaires, is one of the most popular approaches to evaluate health conditions. While a questionnaire consists of as many as 400 questions, people are likely to lose patience. This paper presents a machine learning method to mine the sub-health related questions and then provide classification suggestion based on the self-reporting data collected from Sub-health Condition Identification and Classification Research project. To study the most effective mining approaches, four different feature selection methods were applied to discovery the internal relationship among questions and four different supervised learning classifiers were utilized to investigate the most related questions to the specific diagnostic tasks. Experimental results show that artificial neural network achieves the best performance and the final diagnostic accuracy reaches 84.07% with 20 most related questions.},
bibtype = {inProceedings},
author = {Li Huang, undefined and Shixing Yan, undefined and Jiamin Yuan, undefined and Zhiya Zuo, undefined and Fuping Xu, undefined and Yanzhao Lin, undefined and Mary Qu Yang, undefined and Zhimin Yang, undefined and Li, Guo-Zheng},
booktitle = {2016 IEEE International Conference on Bioinformatics and Biomedicine (BIBM)}
}On the Application of Rough Sets to Skeletal Maturation Classification. Garza-Morales, R., López-Irarragori, F., & Sanchez, R. Artif. Intell. Rev., 45(4):489--508, Kluwer Academic Publishers, Norwell, MA, USA, April, 2016.
Paper doi abstract bibtex
Paper doi abstract bibtex Assessment of skeletal maturation is important for the accurate diagnosis and medical treatment of many disorders and syndromes. However, determining skeletal maturation is not a trivial task, and requires professional medical training. The aim of this paper is to review the application of classification techniques to the problem of identifying the skeletal maturation stage of individuals, in order to provide the specialists a second opinion to backup or reject their assessments. A methodology based on Rough Sets is developed, which formulates skeletal maturation as a multicriteria classification problem, and generates classification rules employing data from lateral radiographs. Our methodology introduces the concept of transition maturity stages to obtain a finer classification on the data. Our empirical evaluation shows that the rules generated match the terms used by experts to determine maturation stage. Furthermore, our rough sets methodology produces the best results in our case of study, both in terms of coverage on the data and accuracy of the classification process, with respect to alternative classification approaches.
@article{Garza-Morales:2016:ARS:2903277.2903297,
author = {Garza-Morales, Rodolfo and L\'{o}pez-Irarragori, Fernando and Sanchez, Romeo},
title = {On the Application of Rough Sets to Skeletal Maturation Classification},
journal = {Artif. Intell. Rev.},
issue_date = {April 2016},
volume = {45},
number = {4},
month = apr,
year = {2016},
issn = {0269-2821},
pages = {489--508},
numpages = {20},
url = {http://dx.doi.org/10.1007/s10462-015-9450-x},
doi = {10.1007/s10462-015-9450-x},
acmid = {2903297},
publisher = {Kluwer Academic Publishers},
address = {Norwell, MA, USA},
abstract="Assessment of skeletal maturation is important for the accurate diagnosis and medical treatment of many disorders and syndromes. However, determining skeletal maturation is not a trivial task, and requires professional medical training. The aim of this paper is to review the application of classification techniques to the problem of identifying the skeletal maturation stage of individuals, in order to provide the specialists a second opinion to backup or reject their assessments. A methodology based on Rough Sets is developed, which formulates skeletal maturation as a multicriteria classification problem, and generates classification rules employing data from lateral radiographs. Our methodology introduces the concept of transition maturity stages to obtain a finer classification on the data. Our empirical evaluation shows that the rules generated match the terms used by experts to determine maturation stage. Furthermore, our rough sets methodology produces the best results in our case of study, both in terms of coverage on the data and accuracy of the classification process, with respect to alternative classification approaches.",
keywords = {Machine learning, Multicriteria classification, Rough sets, Skeletal maturation},
}Variational Autoencoders Explained. August, 2016.
Paper abstract bibtex
Paper abstract bibtex In my previous post about generative adversarial networks, I went over a simple method to training a network that could generate realistic-looking images. However, there were a couple of downsides to using a plain GAN. First, the images are generated off some arbitrary noise. If you wanted to generate a
@misc{noauthor_variational_2016,
title = {Variational {Autoencoders} {Explained}},
url = {http://kvfrans.com/variational-autoencoders-explained/},
abstract = {In my previous post about generative adversarial networks, I went over a simple method to training a network that could generate realistic-looking images. However, there were a couple of downsides to using a plain GAN. First, the images are generated off some arbitrary noise. If you wanted to generate a},
urldate = {2018-05-04TZ},
journal = {kevin frans},
month = aug,
year = {2016},
keywords = {\#Introductory Overview, AI=Artificial Intelligence, Algorithms, Classification, Computer Science, Computer Vision, Data Science, GAN=Generative Adversarial Network, Image Analysis, Informatics, Machine Learning, Machine Learning Applications, Neural Networks, Source Code, Statistics - Machine Learning, Variational Autoencoders, Working Code}
}Segmentation-Free Estimation of Kidney Volumes in CT with Dual Regression Forests. Hussain, A., Hamarneh, G., O'Connell, T., Mohammed, M., & Abugharbieh, R. In Medical Image Computing and Computer-Assisted Intervention Workshop on Machine Learning in Medical Imaging (MICCAI MLMI), volume 10019, pages 156-163, 2016.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai_mlmi2016c,
OPTADDRESS = {},
AUTHOR = {Arafat Hussain and Ghassan Hamarneh and Tim O'Connell and
Mohammed Mohammed and Rafeef Abugharbieh},
BOOKTITLE = {Medical Image Computing and Computer-Assisted Intervention Workshop on Machine Learning in Medical Imaging (MICCAI MLMI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {156-163},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Segmentation-Free Estimation of Kidney Volumes in CT
with Dual Regression Forests},
VOLUME = {10019},
YEAR = {2016},
OPTABSTRACT = {},
DOI = {10.1007/978-3-319-47157-0_19},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Segmentation},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai_mlmi2016c.pdf}
}Hawkes Processes with Stochastic Excitations. Lee, Y., Lim, K. W., & Ong, C. S. In Balcan, M. F. & Weinberger, K. Q., editors, Proceedings of The 33rd International Conference on Machine Learning, volume 48, of Proceedings of Machine Learning Research, pages 79--88, New York, New York, USA, 20--22 Jun, 2016. PMLR.
Paper abstract bibtex
Paper abstract bibtex We propose an extension to Hawkes processes by treating the levels of self-excitation as a stochastic differential equation. Our new point process allows better approximation in application domains where events and intensities accelerate each other with correlated levels of contagion. We generalize a recent algorithm for simulating draws from Hawkes processes whose levels of excitation are stochastic processes, and propose a hybrid Markov chain Monte Carlo approach for model fitting. Our sampling procedure scales linearly with the number of required events and does not require stationarity of the point process. A modular inference procedure consisting of a combination between Gibbs and Metropolis Hastings steps is put forward. We recover expectation maximization as a special case. Our general approach is illustrated for contagion following geometric Brownian motion and exponential Langevin dynamics.
@inproceedings{pmlr-v48-leea16,
Abstract = {We propose an extension to Hawkes processes by treating the levels of self-excitation as a stochastic differential equation. Our new point process allows better approximation in application domains where events and intensities accelerate each other with correlated levels of contagion. We generalize a recent algorithm for simulating draws from Hawkes processes whose levels of excitation are stochastic processes, and propose a hybrid Markov chain Monte Carlo approach for model fitting. Our sampling procedure scales linearly with the number of required events and does not require stationarity of the point process. A modular inference procedure consisting of a combination between Gibbs and Metropolis Hastings steps is put forward. We recover expectation maximization as a special case. Our general approach is illustrated for contagion following geometric Brownian motion and exponential Langevin dynamics.},
Address = {New York, New York, USA},
Author = {Young Lee and Kar Wai Lim and Cheng Soon Ong},
Booktitle = {Proceedings of The 33rd International Conference on Machine Learning},
Date-Modified = {2017-09-27 08:52:57 +0000},
Editor = {Maria Florina Balcan and Kilian Q. Weinberger},
Keywords = {machine learning},
Month = {20--22 Jun},
Pages = {79--88},
Pdf = {http://proceedings.mlr.press/v48/leea16.pdf},
Publisher = {PMLR},
Series = {Proceedings of Machine Learning Research},
Title = {Hawkes Processes with Stochastic Excitations},
Url = {http://proceedings.mlr.press/v48/leea16.html},
Volume = {48},
Year = {2016},
Bdsk-Url-1 = {http://proceedings.mlr.press/v48/leea16.html}}Predictive modeling of colorectal cancer using a dedicated pre-processing pipeline on routine electronic medical records. Kop, R., Hoogendoorn, M., ten Teije , A., Büchner, F., Slottje, P., Moons, L., & Numans, M. Computers in Biology and Medicine, 76:30–38, Elsevier Limited, 9, 2016.
doi abstract bibtex
doi abstract bibtex
Over the past years, research utilizing routine care data extracted from Electronic Medical Records (EMRs) has increased tremendously. Yet there are no straightforward, standardized strategies for pre-processing these data. We propose a dedicated medical pre-processing pipeline aimed at taking on many problems and opportunities contained within EMR data, such as their temporal, inaccurate and incomplete nature. The pipeline is demonstrated on a dataset of routinely recorded data in general practice EMRs of over 260,000 patients, in which the occurrence of colorectal cancer (CRC) is predicted using various machine learning techniques (i.e., CART, LR, RF) and subsets of the data. CRC is a common type of cancer, of which early detection has proven to be important yet challenging. The results are threefold. First, the predictive models generated using our pipeline reconfirmed known predictors and identified new, medically plausible, predictors derived from the cardiovascular and metabolic disease domain, validating the pipeline's effectiveness. Second, the difference between the best model generated by the data-driven subset (AUC 0.891) and the best model generated by the current state of the art hypothesis-driven subset (AUC 0.864) is statistically significant at the 95% confidence interval level. Third, the pipeline itself is highly generic and independent of the specific disease targeted and the EMR used. In conclusion, the application of established machine learning techniques in combination with the proposed pipeline on EMRs has great potential to enhance disease prediction, and hence early detection and intervention in medical practice.
@article{8fcbb631b38f4a52999735c46e47434a,
title = "Predictive modeling of colorectal cancer using a dedicated pre-processing pipeline on routine electronic medical records",
abstract = "Over the past years, research utilizing routine care data extracted from Electronic Medical Records (EMRs) has increased tremendously. Yet there are no straightforward, standardized strategies for pre-processing these data. We propose a dedicated medical pre-processing pipeline aimed at taking on many problems and opportunities contained within EMR data, such as their temporal, inaccurate and incomplete nature. The pipeline is demonstrated on a dataset of routinely recorded data in general practice EMRs of over 260,000 patients, in which the occurrence of colorectal cancer (CRC) is predicted using various machine learning techniques (i.e., CART, LR, RF) and subsets of the data. CRC is a common type of cancer, of which early detection has proven to be important yet challenging. The results are threefold. First, the predictive models generated using our pipeline reconfirmed known predictors and identified new, medically plausible, predictors derived from the cardiovascular and metabolic disease domain, validating the pipeline's effectiveness. Second, the difference between the best model generated by the data-driven subset (AUC 0.891) and the best model generated by the current state of the art hypothesis-driven subset (AUC 0.864) is statistically significant at the 95% confidence interval level. Third, the pipeline itself is highly generic and independent of the specific disease targeted and the EMR used. In conclusion, the application of established machine learning techniques in combination with the proposed pipeline on EMRs has great potential to enhance disease prediction, and hence early detection and intervention in medical practice.",
keywords = "Colorectal cancer, Data mining, Data processing, Electronic medical records, Machine learning",
author = "Reinier Kop and Mark Hoogendoorn and {ten Teije}, Annette and Büchner, {Frederike L.} and Pauline Slottje and Moons, {Leon M G} and Numans, {Mattijs E.}",
year = "2016",
month = "9",
doi = "10.1016/j.compbiomed.2016.06.019",
volume = "76",
pages = "30--38",
journal = "Computers in Biology and Medicine",
issn = "0010-4825",
publisher = "Elsevier Limited",
}2015 (10)
Probabilistic event calculus for event recognition. Skarlatidis, A., Paliouras, G., Artikis, A., & Vouros, G. A. ACM Trans. Comput. Logic, 16(2):11:1--11:37, ACM, New York, NY, USA, feb, 2015.
Paper doi bibtex
Paper doi bibtex @Article{Skarlatidis_2015_17615,
author = {Skarlatidis, Anastasios and Paliouras, Georgios and Artikis, Alexander and Vouros, George A.},
address = {New York, NY, USA},
journal = {ACM Trans. Comput. Logic},
month = {feb},
number = {2},
pages = {11:1--11:37},
publisher = {ACM},
title = {Probabilistic event calculus for event recognition},
volume = {16},
year = {2015},
issn = {1529-3785},
keywords = {Events, machine learning, probabilistic inference, uncertainty},
url = {http://doi.acm.org/10.1145/2699916},
doi = {10.1145/2699916},
title_with_no_special_chars = {Probabilistic Event Calculus for Event Recognition}
}Grand Challenge Veterinary Imaging: Technology, Science, and Communication. McEvoy, F., J. Frontiers in veterinary science, 2:38, 9, 2015.
bibtex
bibtex
@article{
title = {Grand Challenge Veterinary Imaging: Technology, Science, and Communication},
type = {article},
year = {2015},
identifiers = {[object Object]},
keywords = {clinical veterinary radiology,machine learning,teleradiology,translational models,veterinary imaging},
pages = {38},
volume = {2},
month = {9},
day = {30},
city = {Department of Veterinary Clinical and Animal Sciences, University of Copenhagen , Copenhagen , Denmark.},
id = {cb924eb7-534d-3aee-823a-b93631ee0c31},
created = {2016-09-06T13:34:58.000Z},
file_attached = {false},
profile_id = {cacab941-be62-3845-982b-a7700857a11d},
last_modified = {2016-09-07T14:54:40.000Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
source_type = {JOUR},
notes = {JID: 101666658; OTO: NOTNLM; 2015 [ecollection]; 2015/07/27 [received]; 2015/09/07 [accepted]; 2015/09/30 [epublish]; epublish},
bibtype = {article},
author = {McEvoy, F J},
journal = {Frontiers in veterinary science}
}Metadata Dependent Mondrian Processes. Wang, Y., Li, B., Wang, Y., & Chen, F. In Proceedings of the 32Nd International Conference on International Conference on Machine Learning - Volume 37, of ICML'15, pages 1339--1347, 2015. JMLR.org.
Paper bibtex
Paper bibtex @inproceedings{Wang:2015:MDM:3045118.3045261,
Acmid = {3045261},
Author = {Wang, Yi and Li, Bin and Wang, Yang and Chen, Fang},
Booktitle = {Proceedings of the 32Nd International Conference on International Conference on Machine Learning - Volume 37},
Date-Modified = {2017-09-27 09:00:24 +0000},
Keywords = {machine learning},
Location = {Lille, France},
Numpages = {9},
Pages = {1339--1347},
Publisher = {JMLR.org},
Series = {ICML'15},
Title = {Metadata Dependent Mondrian Processes},
Url = {http://dl.acm.org/citation.cfm?id=3045118.3045261},
Year = {2015},
Bdsk-Url-1 = {http://dl.acm.org/citation.cfm?id=3045118.3045261}}Anchored Discrete Factor Analysis. Halpern, Y., Horng, S., & Sontag, D. In arXiv:1511.03299, 2015.
Paper abstract bibtex
Paper abstract bibtex We present a semi-supervised learning algorithm for learning discrete factor analysis models with arbitrary structure on the latent variables. Our algorithm assumes that every latent variable has an "anchor", an observed variable with only that latent variable as its parent. Given such anchors, we show that it is possible to consistently recover moments of the latent variables and use these moments to learn complete models. We also introduce a new technique for improving the robustness of method-of-moment algorithms by optimizing over the marginal polytope or its relaxations. We evaluate our algorithm using two real-world tasks, tag prediction on questions from the Stack Overflow website and medical diagnosis in an emergency department.
@inproceedings{HalpernEtAl_arxiv15,
author = {Yoni Halpern and Steven Horng and David Sontag},
title = {Anchored Discrete Factor Analysis},
booktitle = {arXiv:1511.03299},
year = {2015},
keywords = {Machine learning, Unsupervised learning, Health care},
url_Paper = {http://arxiv.org/pdf/1511.03299.pdf},
abstract = {We present a semi-supervised learning algorithm for learning discrete factor analysis models with arbitrary structure on the latent variables. Our algorithm assumes that every latent variable has an "anchor", an observed variable with only that latent variable as its parent. Given such anchors, we show that it is possible to consistently recover moments of the latent variables and use these moments to learn complete models. We also introduce a new technique for improving the robustness of method-of-moment algorithms by optimizing over the marginal polytope or its relaxations. We evaluate our algorithm using two real-world tasks, tag prediction on questions from the Stack Overflow website and medical diagnosis in an emergency department.}
}Fine-grained OD estimation with automated zoning and sparsity regularisation. Menon, A. K., Cai, C., Wang, W., Wen, T., & Chen, F. Transportation Research Part B: Methodological, 80:150--172, Elsevier BV, oct, 2015.
Paper doi bibtex
Paper doi bibtex @article{Menon_2015,
Author = {Aditya Krishna Menon and Chen Cai and Weihong Wang and Tao Wen and Fang Chen},
Date-Modified = {2017-09-27 08:53:52 +0000},
Doi = {10.1016/j.trb.2015.07.003},
Journal = {Transportation Research Part B: Methodological},
Keywords = {machine learning},
Month = {oct},
Pages = {150--172},
Publisher = {Elsevier {BV}},
Title = {Fine-grained {OD} estimation with automated zoning and sparsity regularisation},
Url = {https://doi.org/10.1016%2Fj.trb.2015.07.003},
Volume = {80},
Year = 2015,
Bdsk-Url-1 = {https://doi.org/10.1016%2Fj.trb.2015.07.003},
Bdsk-Url-2 = {http://dx.doi.org/10.1016/j.trb.2015.07.003}}A computer vision tracking system for pigmented skin lesions. Mirzaalian, H., Lee, T., & Hamarneh, G. In World Congress of Dermatology (WCD), 2015.
bibtex
bibtex
@INPROCEEDINGS{wcd2015b,
OPTADDRESS = {},
AUTHOR = {Hengameh Mirzaalian and Tim Lee and Ghassan Hamarneh},
BOOKTITLE = {World Congress of Dermatology (WCD)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
OPTPAGES = {},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {A computer vision tracking system for pigmented skin
lesions},
OPTVOLUME = {},
YEAR = {2015},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Processing, Color, Machine Learning, Optimization,
Registration and Matching, Tracking, Graph based, Dermatology},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/wcd2015b.pdf}
}Cross-Modal Retrieval: A Pairwise Classification Approach. Menon, A. K., Surian, D., & Chawla, S. In Proceedings of the 2015 SIAM International Conference on Data Mining, pages 199--207. Society for Industrial and Applied Mathematics, jun, 2015.
Paper doi bibtex
Paper doi bibtex @incollection{Menon_2015b,
Author = {Aditya Krishna Menon and Didi Surian and Sanjay Chawla},
Booktitle = {Proceedings of the 2015 {SIAM} International Conference on Data Mining},
Date-Modified = {2017-09-27 08:53:52 +0000},
Doi = {10.1137/1.9781611974010.23},
Keywords = {machine learning},
Month = {jun},
Pages = {199--207},
Publisher = {Society for Industrial and Applied Mathematics},
Title = {Cross-Modal Retrieval: A Pairwise Classification Approach},
Url = {https://doi.org/10.1137%2F1.9781611974010.23},
Year = 2015,
Bdsk-Url-1 = {https://doi.org/10.1137%2F1.9781611974010.23},
Bdsk-Url-2 = {http://dx.doi.org/10.1137/1.9781611974010.23}}Detecting Streaks from Dermoscopic Images of Pigmented Skin Lesions. Mirzaalian, H., Lee, T., & Hamarneh, G. In The International Society for Biophysics and Imaging of the Skin (ISBS), 2015.
bibtex
bibtex
@INPROCEEDINGS{isbs2015,
OPTADDRESS = {},
AUTHOR = {Hengameh Mirzaalian and Tim Lee and Ghassan Hamarneh},
BOOKTITLE = {The International Society for Biophysics and Imaging of the Skin (ISBS)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
OPTPAGES = {},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Detecting Streaks from Dermoscopic Images of Pigmented
Skin Lesions},
OPTVOLUME = {},
YEAR = {2015},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Processing, Color, Machine Learning, Dermatology},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/isbs2015.pdf}
}Machine learning applications in cancer prognosis and prediction. Kourou, K., Exarchos, T., P., Exarchos, K., P., Karamouzis, M., V., & Fotiadis, D., I. Computational and Structural Biotechnology Journal, 13:8-17, 12, 2015.
Website doi abstract bibtex
Website doi abstract bibtex Cancer has been characterized as a heterogeneous disease consisting of many different subtypes. The early diagnosis and prognosis of a cancer type have become a necessity in cancer research, as it can facilitate the subsequent clinical management of patients. The importance of classifying cancer patients into high or low risk groups has led many research teams, from the biomedical and the bioinformatics field, to study the application of machine learning (ML) methods. Therefore, these techniques have been utilized as an aim to model the progression and treatment of cancerous conditions. In addition, the ability of ML tools to detect key features from complex datasets reveals their importance. A variety of these techniques, including Artificial Neural Networks (ANNs), Bayesian Networks (BNs), Support Vector Machines (SVMs) and Decision Trees (DTs) have been widely applied in cancer research for the development of predictive models, resulting in effective and accurate decision making. Even though it is evident that the use of ML methods can improve our understanding of cancer progression, an appropriate level of validation is needed in order for these methods to be considered in the everyday clinical practice. In this work, we present a review of recent ML approaches employed in the modeling of cancer progression. The predictive models discussed here are based on various supervised ML techniques as well as on different input features and data samples. Given the growing trend on the application of ML methods in cancer research, we present here the most recent publications that employ these techniques as an aim to model cancer risk or patient outcomes.
@article{
title = {Machine learning applications in cancer prognosis and prediction},
type = {article},
year = {2015},
keywords = {ANN,AUC,Area Under Curve,Artificial Neural Network,BCRSVM,BN,Bayesian Network,Breast Cancer Support Vector Machine,CFS,Cancer recurrence,Cancer survival,Cancer susceptibility,Correlation based Feature Selection,DT,Decision Tree,ES,Early Stopping algorithm,GEO,Gene Expression Omnibus,HTT,High-throughput Technologies,LCS,Learning Classifying Systems,ML,Machine Learning,Machine learning,NCI caArray,NSCLC,National Cancer Institute Array Data Management Sy,Non-small Cell Lung Cancer,OSCC,Oral Squamous Cell Carcinoma,PPI,Predictive models,Protein–Protein Interaction,ROC,Receiver Operating Characteristic,SEER,SSL,SVM,Semi-supervised Learning,Support Vector Machine,Surveillance, Epidemiology and End results Databas,TCGA,The Cancer Genome Atlas Research Network},
pages = {8-17},
volume = {13},
websites = {http://www.sciencedirect.com/science/article/pii/S2001037014000464},
month = {12},
id = {4366c512-fb41-38bf-b53a-aa5d57feec95},
created = {2015-04-12T18:59:39.000Z},
accessed = {2015-01-01},
file_attached = {false},
profile_id = {95e10851-cdf3-31de-9f82-1ab629e601b0},
group_id = {838ecfe2-7c01-38b2-970d-875a87910530},
last_modified = {2017-03-14T14:27:28.880Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
private_publication = {false},
abstract = {Cancer has been characterized as a heterogeneous disease consisting of many different subtypes. The early diagnosis and prognosis of a cancer type have become a necessity in cancer research, as it can facilitate the subsequent clinical management of patients. The importance of classifying cancer patients into high or low risk groups has led many research teams, from the biomedical and the bioinformatics field, to study the application of machine learning (ML) methods. Therefore, these techniques have been utilized as an aim to model the progression and treatment of cancerous conditions. In addition, the ability of ML tools to detect key features from complex datasets reveals their importance. A variety of these techniques, including Artificial Neural Networks (ANNs), Bayesian Networks (BNs), Support Vector Machines (SVMs) and Decision Trees (DTs) have been widely applied in cancer research for the development of predictive models, resulting in effective and accurate decision making. Even though it is evident that the use of ML methods can improve our understanding of cancer progression, an appropriate level of validation is needed in order for these methods to be considered in the everyday clinical practice. In this work, we present a review of recent ML approaches employed in the modeling of cancer progression. The predictive models discussed here are based on various supervised ML techniques as well as on different input features and data samples. Given the growing trend on the application of ML methods in cancer research, we present here the most recent publications that employ these techniques as an aim to model cancer risk or patient outcomes.},
bibtype = {article},
author = {Kourou, Konstantina and Exarchos, Themis P. and Exarchos, Konstantinos P. and Karamouzis, Michalis V. and Fotiadis, Dimitrios I.},
doi = {10.1016/j.csbj.2014.11.005},
journal = {Computational and Structural Biotechnology Journal}
}Identifying Relevant Messages in a Twitter-based Citizen Channel for Natural Disaster Situations. Cobo, A., Parra, D., & Navón, J. In Proceedings of the 24th International Conference on World Wide Web, of WWW '15 Companion, pages 1189–1194, Republic and Canton of Geneva, Switzerland, 2015. International World Wide Web Conferences Steering Committee.
Paper doi bibtex
Paper doi bibtex @inproceedings{Cobo:2015:IRM:2740908.2741719,
Acmid = {2741719},
Address = {Republic and Canton of Geneva, Switzerland},
Author = {Cobo, Alfredo and Parra, Denis and Nav\'{o}n, Jaime},
Booktitle = {Proceedings of the 24th International Conference on World Wide Web},
Date-Added = {2015-12-30 15:21:21 +0000},
Date-Modified = {2015-12-30 15:24:32 +0000},
Doi = {10.1145/2740908.2741719},
Isbn = {978-1-4503-3473-0},
Keywords = {class imbalance, machine learning, natural disaster, twitter},
Location = {Florence, Italy},
Numpages = {6},
Pages = {1189--1194},
Publisher = {International World Wide Web Conferences Steering Committee},
Series = {WWW '15 Companion},
Title = {Identifying Relevant Messages in a Twitter-based Citizen Channel for Natural Disaster Situations},
Year = {2015},
url = {http://web.ing.puc.cl/~dparra/pdfs/Cobo20151503.05784v1.pdf},
Bdsk-Url-1 = {http://dx.doi.org/10.1145/2740908.2741719}}2014 (8)
ARTgrid: A Two-Level Learning Architecture Based on Adaptive Resonance Theory. Švaco, M. & Jerbić, B. Advances in Artificial Neural Systems, 2014:1–9, 2014.
doi bibtex
doi bibtex
@article{svaco_artgrid_2014,
title = {{ARTgrid}: {A} {Two}-{Level} {Learning} {Architecture} {Based} on {Adaptive} {Resonance} {Theory}},
volume = {2014},
doi = {10.1155/2014/185492},
journal = {Advances in Artificial Neural Systems},
author = {Švaco, Marko and Jerbić, Bojan},
year = {2014},
keywords = {ARTgrid, Adaptive resonance theory, Machine learning},
pages = {1--9}
}Software Bug Localization with Markov Logic. Zhang, S. 2014.
Paper bibtex
Paper bibtex @article{
title = {Software Bug Localization with Markov Logic},
type = {article},
year = {2014},
identifiers = {[object Object]},
keywords = {automated debugging,machine learning},
pages = {424-427},
id = {c09814f3-a934-307a-b553-560c7b8d0b8a},
created = {2014-07-22T07:03:13.000Z},
file_attached = {true},
profile_id = {6670f279-692b-3eed-bc36-b89af19ad6e0},
group_id = {272c437f-364d-3eb7-83a6-2112aad7b3e7},
last_modified = {2014-07-22T09:25:04.000Z},
tags = {new_ideas},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
citation_key = {Zhang2014},
bibtype = {article},
author = {Zhang, Sai}
}A machine learning method for high-frequency data forecasting. Allende, H., López, E., & Allende-Cid, H. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), volume 8827, pages 621-628, 2014.
abstract bibtex
abstract bibtex
© Springer International Publishing Switzerland 2014.In recent years several models for financial high-frequency data have been proposed. One of the most known models for this type of applications is the ACM-ACD model. This model focuses on modelling the underlying joint distribution of both duration and price changes between consecutive transactions. However this model imposes distributional assumptions and its number of parameters increases rapidly (producing a complex and slow adjustment process). Therefore, we propose using two machine learning models, that will work sequentially, based on the ACM-ACD model. The results show a comparable performance, achieving a better performance in some cases. Also the proposal achieves a significatively more rapid convergence. The proposal is validated with a well-known financial data set.
@inproceedings{84949124234,
abstract = "© Springer International Publishing Switzerland 2014.In recent years several models for financial high-frequency data have been proposed. One of the most known models for this type of applications is the ACM-ACD model. This model focuses on modelling the underlying joint distribution of both duration and price changes between consecutive transactions. However this model imposes distributional assumptions and its number of parameters increases rapidly (producing a complex and slow adjustment process). Therefore, we propose using two machine learning models, that will work sequentially, based on the ACM-ACD model. The results show a comparable performance, achieving a better performance in some cases. Also the proposal achieves a significatively more rapid convergence. The proposal is validated with a well-known financial data set.",
year = "2014",
title = "A machine learning method for high-frequency data forecasting",
volume = "8827",
keywords = "ACM-ACD model , Financial high-frequency data , Forecasting , Machine learning , Time series",
pages = "621-628",
booktitle = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
author = "Allende, Héctor and López, Erick and Allende-Cid, Héctor"
}Ventricular fibrillation and tachycardia classification using a machine learning approach. Li, Q., Rajagopalan, C., & Clifford, G., D. IEEE Transactions on Biomedical Engineering, 61(6):1607-1613, IEEE Computer Society, 2014.
abstract bibtex
abstract bibtex
Correct detection and classification of ventricular fibrillation (VF) and rapid ventricular tachycardia (VT) is of pivotal importance for an automatic external defibrillator and patient monitoring. In this paper, a VF/VT classification algorithm using a machine learning method, a support vector machine, is proposed. A total of 14 metrics were extracted from a specific window length of the electrocardiogram (ECG). A genetic algorithm was then used to select the optimal variable combinations. Three annotated public domain ECG databases (the American Heart Association Database, the Creighton University Ventricular Tachyarrhythmia Database, and the MIT-BIH Malignant Ventricular Arrhythmia Database) were used as training, test, and validation datasets. Different window sizes, varying from 1 to 10 s were tested. An accuracy (Ac) of 98.1%, sensitivity (Se) of 98.4%, and specificity (Sp) of 98.0% were obtained on the in-sample training data with 5 s-window size and two selected metrics. On the out-of-sample validation data, an Ac of 96.3% ± 3.4%, Se of 96.2% ± 2.7%, and Sp of 96.2% ± 4.6% were obtained by fivefold cross validation. The results surpass those of current reported methods.
@article{
title = {Ventricular fibrillation and tachycardia classification using a machine learning approach},
type = {article},
year = {2014},
identifiers = {[object Object]},
keywords = {Machine learning,public domain electrocardiogram (ECG) database,support vector machine (SVM),ventricular fibrillation (VF) detection},
pages = {1607-1613},
volume = {61},
publisher = {IEEE Computer Society},
id = {2ea42df9-e5ec-3806-bdbf-082f6040cf1a},
created = {2016-03-29T18:26:51.000Z},
file_attached = {false},
profile_id = {304786e8-5116-360a-80be-e62833097578},
group_id = {d7b44578-07c1-3210-ae74-3bcd7f980767},
last_modified = {2017-03-14T15:45:25.917Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
citation_key = {Li2014},
private_publication = {false},
abstract = {Correct detection and classification of ventricular fibrillation (VF) and rapid ventricular tachycardia (VT) is of pivotal importance for an automatic external defibrillator and patient monitoring. In this paper, a VF/VT classification algorithm using a machine learning method, a support vector machine, is proposed. A total of 14 metrics were extracted from a specific window length of the electrocardiogram (ECG). A genetic algorithm was then used to select the optimal variable combinations. Three annotated public domain ECG databases (the American Heart Association Database, the Creighton University Ventricular Tachyarrhythmia Database, and the MIT-BIH Malignant Ventricular Arrhythmia Database) were used as training, test, and validation datasets. Different window sizes, varying from 1 to 10 s were tested. An accuracy (Ac) of 98.1%, sensitivity (Se) of 98.4%, and specificity (Sp) of 98.0% were obtained on the in-sample training data with 5 s-window size and two selected metrics. On the out-of-sample validation data, an Ac of 96.3% ± 3.4%, Se of 96.2% ± 2.7%, and Sp of 96.2% ± 4.6% were obtained by fivefold cross validation. The results surpass those of current reported methods.},
bibtype = {article},
author = {Li, Qiao and Rajagopalan, Cadathur and Clifford, Gari D.},
journal = {IEEE Transactions on Biomedical Engineering},
number = {6}
}Making Big Data Useful for Health Care: A Summary of the Inaugural MIT Critical Data Conference. Badawi, O., Brennan, T., Celi, L. A., Feng, M., Ghassemi, M., Ippolito, A., Johnson, A., Mark, R. G, Mayaud, L., Moody, G., Moses, C., Naumann, T., Nikore, V., Pimentel, M., Pollard, T. J, Santos, M., Stone, D. J, Zimolzak, A., & MIT Critical Data Conference 2014 Organizing Committee JMIR Medical Informatics, 2(2):e22, August, 2014.
Paper doi bibtex
Paper doi bibtex @article{badawi_making_2014,
title = {Making {Big} {Data} {Useful} for {Health} {Care}: {A} {Summary} of the {Inaugural} {MIT} {Critical} {Data} {Conference}},
volume = {2},
issn = {2291-9694},
shorttitle = {Making {Big} {Data} {Useful} for {Health} {Care}},
url = {http://medinform.jmir.org/2014/2/e22/},
doi = {10.2196/medinform.3447},
language = {en},
number = {2},
urldate = {2017-08-14TZ},
journal = {JMIR Medical Informatics},
author = {Badawi, Omar and Brennan, Thomas and Celi, Leo Anthony and Feng, Mengling and Ghassemi, Marzyeh and Ippolito, Andrea and Johnson, Alistair and Mark, Roger G and Mayaud, Louis and Moody, George and Moses, Christopher and Naumann, Tristan and Nikore, Vipan and Pimentel, Marco and Pollard, Tom J and Santos, Mauro and Stone, David J and Zimolzak, Andrew and {MIT Critical Data Conference 2014 Organizing Committee}},
month = aug,
year = {2014},
pages = {e22}
}Harnessing the power of big data: infusing the scientific method with machine learning to transform ecology. Peters, D. P. C., Havstad, K. M., Cushing, J., Tweedie, C., Fuentes, O., & Vilanueva-Rosales, N. Ecosphere, 5(6):67. http://dx.doi.org/10.1890/ES13–00359.1, 2014.
Paper abstract bibtex
Paper abstract bibtex Most efforts to harness the power of big data for ecology and environmental sciences focus on data and metadata sharing, standardization, and accuracy. However, many scientists have not accepted the data deluge as an integral part of their research because the current scientific method is not scalable to large, complex datasets. Here, we explain how integrating a data-intensive, machine learning approach with a hypothesis-driven, mechanistic approach can lead to a novel knowledge, learning, analysis system (KLAS) for discovery and problem solving. Machine learning leads to more efficient, user-friendly analytics as the streams of data increase while hypothesis-driven decisions lead to the strategic design of experiments to fill knowledge gaps and to elucidate mechanisms. KLAS will transform ecology and environmental sciences by shortening the time lag between individual discoveries and leaps in knowledge by the scientific community, and will lead to paradigm shifts predicated on open access data and analytics in a machine learning environment.
@article{peters_harnessing_2014,
title = {Harnessing the power of big data: infusing the scientific method with machine learning to transform ecology},
volume = {5},
url = {http://dx.doi.org/10.1890/ES13-00359.1},
abstract = {Most efforts to harness the power of big data for ecology and environmental sciences focus on data and metadata sharing, standardization, and accuracy. However, many scientists have not accepted the data deluge as an integral part of their research because the current scientific method is not scalable to large, complex datasets. Here, we explain how integrating a data-intensive, machine learning approach with a hypothesis-driven, mechanistic approach can lead to a novel knowledge, learning, analysis system (KLAS) for discovery and problem solving. Machine learning leads to more efficient, user-friendly analytics as the streams of data increase while hypothesis-driven decisions lead to the strategic design of experiments to fill knowledge gaps and to elucidate mechanisms. KLAS will transform ecology and environmental sciences by shortening the time lag between individual discoveries and leaps in knowledge by the scientific community, and will lead to paradigm shifts predicated on open access data and analytics in a machine learning environment.},
number = {6},
journal = {Ecosphere},
author = {Peters, Debra P. C. and Havstad, Kris M. and Cushing, Judy and Tweedie, Craig and Fuentes, Olac and Vilanueva-Rosales, Natalia},
year = {2014},
keywords = {LTER, analytics, article, data deluge, journal, long-term data, machine learning, open data, paradigm shifts},
pages = {67. http://dx.doi.org/10.1890/ES13--00359.1}
}A methodology for the characterization and diagnosis of cognitive impairments-Application to specific language impairment. Oliva, J., Serrano, J., I., del Castillo, M., D., & Iglesias, Á. Artificial Intelligence in Medicine, 61(2):89-96, Elsevier B.V., 2014.
Paper Website abstract bibtex
Paper Website abstract bibtex Objectives: The diagnosis of mental disorders is in most cases very difficult because of the high heterogeneity and overlap between associated cognitive impairments. Furthermore, early and individualized diagnosis is crucial. In this paper, we propose a methodology to support the individualized characterization and diagnosis of cognitive impairments. The methodology can also be used as a test platform for existing theories on the causes of the impairments. We use computational cognitive modeling to gather information on the cognitive mechanisms underlying normal and impaired behavior. We then use this information to feed machine-learning algorithms to individually characterize the impairment and to differentiate between normal and impaired behavior. We apply the methodology to the particular case of specific language impairment (SLI) in Spanish-speaking children. Methods and materials: The proposed methodology begins by defining a task in which normal and individuals with impairment present behavioral differences. Next we build a computational cognitive model of that task and individualize it: we build a cognitive model for each participant and optimize its parameter values to fit the behavior of each participant. Finally, we use the optimized parameter values to feed different machine learning algorithms. The methodology was applied to an existing database of 48 Spanish-speaking children (24 normal and 24 SLI children) using clustering techniques for the characterization, and different classifier techniques for the diagnosis. Results: The characterization results show three well-differentiated groups that can be associated with the three main theories on SLI. Using a leave-one-subject-out testing methodology, all the classifiers except the DT produced sensitivity, specificity and area under curve values above 90%, reaching 100% in some cases. Conclusions: The results show that our methodology is able to find relevant information on the underlying cognitive mechanisms and to use it appropriately to provide better diagnosis than existing techniques. It is also worth noting that the individualized characterization obtained using our methodology could be extremely helpful in designing individualized therapies. Moreover, the proposed methodology could be easily extended to other languages and even to other cognitive impairments not necessarily related to language. © 2014 Elsevier B.V.
@article{
title = {A methodology for the characterization and diagnosis of cognitive impairments-Application to specific language impairment},
type = {article},
year = {2014},
identifiers = {[object Object]},
keywords = {Automatic diagnosis,Computational cognitive modeling,Machine learning,Specific language impairment},
pages = {89-96},
volume = {61},
websites = {http://dx.doi.org/10.1016/j.artmed.2014.04.002},
publisher = {Elsevier B.V.},
id = {bea1cbb4-4a9b-3106-883c-4f2cbd39b65d},
created = {2015-04-29T10:51:20.000Z},
file_attached = {true},
profile_id = {83be13a1-4e81-3e4b-9603-d27f26bebe2e},
group_id = {5ac1deb5-92b2-3f98-87d8-03b81a0fdb16},
last_modified = {2017-03-14T11:09:31.719Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
citation_key = {Oliva2014},
folder_uuids = {d944e9de-e3e6-4e8d-b73f-c1bf50a78ebe},
private_publication = {false},
abstract = {Objectives: The diagnosis of mental disorders is in most cases very difficult because of the high heterogeneity and overlap between associated cognitive impairments. Furthermore, early and individualized diagnosis is crucial. In this paper, we propose a methodology to support the individualized characterization and diagnosis of cognitive impairments. The methodology can also be used as a test platform for existing theories on the causes of the impairments. We use computational cognitive modeling to gather information on the cognitive mechanisms underlying normal and impaired behavior. We then use this information to feed machine-learning algorithms to individually characterize the impairment and to differentiate between normal and impaired behavior. We apply the methodology to the particular case of specific language impairment (SLI) in Spanish-speaking children. Methods and materials: The proposed methodology begins by defining a task in which normal and individuals with impairment present behavioral differences. Next we build a computational cognitive model of that task and individualize it: we build a cognitive model for each participant and optimize its parameter values to fit the behavior of each participant. Finally, we use the optimized parameter values to feed different machine learning algorithms. The methodology was applied to an existing database of 48 Spanish-speaking children (24 normal and 24 SLI children) using clustering techniques for the characterization, and different classifier techniques for the diagnosis. Results: The characterization results show three well-differentiated groups that can be associated with the three main theories on SLI. Using a leave-one-subject-out testing methodology, all the classifiers except the DT produced sensitivity, specificity and area under curve values above 90%, reaching 100% in some cases. Conclusions: The results show that our methodology is able to find relevant information on the underlying cognitive mechanisms and to use it appropriately to provide better diagnosis than existing techniques. It is also worth noting that the individualized characterization obtained using our methodology could be extremely helpful in designing individualized therapies. Moreover, the proposed methodology could be easily extended to other languages and even to other cognitive impairments not necessarily related to language. © 2014 Elsevier B.V.},
bibtype = {article},
author = {Oliva, Jesús and Serrano, J. Ignacio and del Castillo, M. Dolores and Iglesias, Ángel},
journal = {Artificial Intelligence in Medicine},
number = {2}
}Robot Assisted 3D Point Cloud Object Registration. Jerbić, B., Šuligoj, F., Švaco, M., & Šekoranja, B. 2014.
bibtex
bibtex
@article{jerbic_robot_2014,
title = {Robot {Assisted} {3D} {Point} {Cloud} {Object} {Registration}},
author = {Jerbić, Bojan and Šuligoj, Filip and Švaco, Marko and Šekoranja, Bojan},
year = {2014},
keywords = {Point cloud data, machine learning, object recognition, position estimation, stereovision},
pages = {847--852}
}2013 (8)
Decision Forests with Spatio-temporal Features for Graph-based Tumour Segmentation in 4D Lung CT. Mirzaei, H., Tang, L. Y. W., Werner, R., & Hamarneh, G. In Medical Image Computing and Computer-Assisted Intervention Workshop on Machine Learning in Medical Imaging (MICCAI MLMI), volume 8184, pages 179-186, 2013.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai_mlmi2013c,
OPTADDRESS = {},
AUTHOR = {Hamidreza Mirzaei and Lisa Y. W. Tang and Rene Werner and
Ghassan Hamarneh},
BOOKTITLE = {Medical Image Computing and Computer-Assisted Intervention Workshop on Machine Learning in Medical Imaging (MICCAI MLMI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {179-186},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Decision Forests with Spatio-temporal Features for
Graph-based Tumour Segmentation in 4D Lung CT},
VOLUME = {8184},
YEAR = {2013},
OPTABSTRACT = {},
DOI = {10.1007/978-3-319-02267-3_23},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Segmentation, Machine Learning,
Functional/Molecular/Dynamic Imaging},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai_mlmi2013c.pdf}
}Detecting inappropriate access to electronic health records using collaborative filtering. Menon, A. K., Jiang, X., Kim, J., Vaidya, J., & Ohno-Machado, L. Machine Learning, 95(1):87--101, Springer Nature, jun, 2013.
Paper doi bibtex
Paper doi bibtex @article{Menon_2013,
Author = {Aditya Krishna Menon and Xiaoqian Jiang and Jihoon Kim and Jaideep Vaidya and Lucila Ohno-Machado},
Date-Modified = {2017-09-27 08:53:52 +0000},
Doi = {10.1007/s10994-013-5376-1},
Journal = {Machine Learning},
Keywords = {machine learning},
Month = {jun},
Number = {1},
Pages = {87--101},
Publisher = {Springer Nature},
Title = {Detecting inappropriate access to electronic health records using collaborative filtering},
Url = {https://doi.org/10.1007%2Fs10994-013-5376-1},
Volume = {95},
Year = 2013,
Bdsk-Url-1 = {https://doi.org/10.1007%2Fs10994-013-5376-1},
Bdsk-Url-2 = {http://dx.doi.org/10.1007/s10994-013-5376-1}}Parallel globally optimal structure learning of Bayesian networks. Nikolova, O., Zola, J., & Aluru, S. Journal of Parallel and Distributed Computing, 73(8):1039-1048, 8, 2013.
Website doi abstract bibtex
Website doi abstract bibtex Given n random variables and a set of m observations of each of the n variables, the Bayesian network structure learning problem is to learn a directed acyclic graph (DAG) on the n variables such that the implied joint probability distribution best explains the set of observations. Bayesian networks are widely used in many fields including data mining and computational biology. Globally optimal (exact) structure learning of Bayesian networks takes O(n2⋅2n) time plus the cost of O(n⋅2n) evaluations of an application-specific scoring function whose run-time is at least linear in m. In this paper, we present a parallel algorithm for exact structure learning of a Bayesian network that is communication-efficient and work-optimal up to O(1n⋅2n) processors. We further extend this algorithm to the important restricted case of structure learning with bounded node in-degree and investigate the performance gains achievable because of limiting node in-degree. We demonstrate the applicability of our method by implementation on an IBM Blue Gene/P system and an AMD Opteron InfiniBand cluster and present experimental results that characterize run-time behavior with respect to the number of variables, number of observations, and the bound on in-degree.
@article{
title = {Parallel globally optimal structure learning of Bayesian networks},
type = {article},
year = {2013},
keywords = {Bayesian networks,Graphical models,Machine learning,Parallel algorithm,Structure learning},
pages = {1039-1048},
volume = {73},
websites = {http://www.sciencedirect.com/science/article/pii/S0743731513000622},
month = {8},
id = {8fd919c6-2590-3c59-bb3d-e1c236644e79},
created = {2015-04-11T19:52:01.000Z},
accessed = {2015-02-18},
file_attached = {false},
profile_id = {95e10851-cdf3-31de-9f82-1ab629e601b0},
group_id = {4a6a1914-6ba6-3cdc-b1f4-f10a6e56cb6c},
last_modified = {2017-03-14T14:27:43.598Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
private_publication = {false},
abstract = {Given n random variables and a set of m observations of each of the n variables, the Bayesian network structure learning problem is to learn a directed acyclic graph (DAG) on the n variables such that the implied joint probability distribution best explains the set of observations. Bayesian networks are widely used in many fields including data mining and computational biology. Globally optimal (exact) structure learning of Bayesian networks takes O(n2⋅2n) time plus the cost of O(n⋅2n) evaluations of an application-specific scoring function whose run-time is at least linear in m. In this paper, we present a parallel algorithm for exact structure learning of a Bayesian network that is communication-efficient and work-optimal up to O(1n⋅2n) processors. We further extend this algorithm to the important restricted case of structure learning with bounded node in-degree and investigate the performance gains achievable because of limiting node in-degree. We demonstrate the applicability of our method by implementation on an IBM Blue Gene/P system and an AMD Opteron InfiniBand cluster and present experimental results that characterize run-time behavior with respect to the number of variables, number of observations, and the bound on in-degree.},
bibtype = {article},
author = {Nikolova, Olga and Zola, Jaroslaw and Aluru, Srinivas},
doi = {10.1016/j.jpdc.2013.04.001},
journal = {Journal of Parallel and Distributed Computing},
number = {8}
}Online ridge regression method using sliding windows. Arce, P. & Salinas, L. C. In pages 87-90, 2013.
doi abstract bibtex
doi abstract bibtex
A new regression method based on the aggregating algorithm for regression (AAR) is presented. The proposal shows how ridge regression can be modified in order to reduce the number of operations by avoiding the inverse matrix calculation only considering a sliding window of the last input values. This modification allows algorithm expression in a recursive way and therefore its use in an online context. Ridge regression, AAR and our proposal were compared using the closing stock prices of 45 stocks from the technology market from 2000 to 2012. Empirical results show that our proposal performs better than the other two methods in 28 of 45 stocks analyzed, due to the lower MSE error. © 2013 IEEE.
@inproceedings{10.1109/SCCC.2012.18,
abstract = "A new regression method based on the aggregating algorithm for regression (AAR) is presented. The proposal shows how ridge regression can be modified in order to reduce the number of operations by avoiding the inverse matrix calculation only considering a sliding window of the last input values. This modification allows algorithm expression in a recursive way and therefore its use in an online context. Ridge regression, AAR and our proposal were compared using the closing stock prices of 45 stocks from the technology market from 2000 to 2012. Empirical results show that our proposal performs better than the other two methods in 28 of 45 stocks analyzed, due to the lower MSE error. © 2013 IEEE.",
year = "2013",
title = "Online ridge regression method using sliding windows",
keywords = "Machine learning , Online learning , Ridge regression",
pages = "87-90",
doi = "10.1109/SCCC.2012.18",
journal = "Proceedings - International Conference of the Chilean Computer Science Society, SCCC",
author = "Arce, Paola and Salinas, Luís C."
}An Introduction to Statistical Learning with Applications in R. James, G., Witten, D., Hastie, T., & Tibshirani, R. Volume 103 of Springer Texts in Statistics, Springer New York, New York, NY, 2013.
Paper doi bibtex
Paper doi bibtex @book{james_introduction_2013,
address = {New York, NY},
series = {Springer {Texts} in {Statistics}},
title = {An {Introduction} to {Statistical} {Learning} with {Applications} in {R}},
volume = {103},
isbn = {978-1-4614-7137-0 978-1-4614-7138-7},
shorttitle = {An {Introduction} to {Statistical} {Learning}},
url = {http://link.springer.com/10.1007/978-1-4614-7138-7},
urldate = {2021-07-06},
publisher = {Springer New York},
author = {James, Gareth and Witten, Daniela and Hastie, Trevor and Tibshirani, Robert},
year = {2013},
doi = {10.1007/978-1-4614-7138-7},
keywords = {ML, Machine Learning, R-project},
}Unsupervised Learning of Noisy-Or Bayesian Networks. Halpern, Y. & Sontag, D. In Proceedings of the Twenty-Ninth Conference on Uncertainty in Artificial Intelligence (UAI-13), pages 272–281, Corvallis, Oregon, 2013. AUAI Press.
Paper abstract bibtex
Paper abstract bibtex This paper considers the problem of learning the parameters in Bayesian networks of discrete variables with known structure and hidden variables. Previous approaches in these settings typically use expectation maximization; when the network has high treewidth, the required expectations might be approximated using Monte Carlo or variational methods. We show how to avoid inference altogether during learning by giving a polynomial-time algorithm based on the method-of-moments, building upon recent work on learning discrete-valued mixture models. In particular, we show how to learn the parameters for a family of bipartite noisy-or Bayesian networks. In our experimental results, we demonstrate an application of our algorithm to learning QMR-DT, a large Bayesian network used for medical diagnosis. We show that it is possible to fully learn the parameters of QMR-DT even when only the findings are observed in the training data (ground truth diseases unknown).
@inproceedings{HalpernSontag_uai13,
author = {Yoni Halpern and David Sontag},
title = {Unsupervised Learning of Noisy-Or Bayesian Networks},
booktitle = {Proceedings of the Twenty-Ninth Conference on Uncertainty in Artificial Intelligence ({UAI}-13)},
publisher = {AUAI Press},
address = {Corvallis, Oregon},
pages = {272--281},
year = {2013},
keywords = {Machine learning, Unsupervised learning, Health care},
url_Paper = {http://people.csail.mit.edu/dsontag/papers/HalpernSontag_uai13.pdf},
abstract = {This paper considers the problem of learning the parameters in Bayesian networks of discrete variables with known structure and hidden variables. Previous approaches in these settings typically use expectation maximization; when the network has high treewidth, the required expectations might be approximated using Monte Carlo or variational methods. We show how to avoid inference altogether during learning by giving a polynomial-time algorithm based on the method-of-moments, building upon recent work on learning discrete-valued mixture models. In particular, we show how to learn the parameters for a family of bipartite noisy-or Bayesian networks. In our experimental results, we demonstrate an application of our algorithm to learning QMR-DT, a large Bayesian network used for medical diagnosis. We show that it is possible to fully learn the parameters of QMR-DT even when only the findings are observed in the training data (ground truth diseases unknown).}
}Novel Morphological and Appearance Features for Predicting Physical Disability from MR Images in Multiple Sclerosis Patients. Kawahara, J., McIntosh, C., Tam, R., & Hamarneh, G. In Medical Image Computing and Computer-Assisted Intervention Workshop on Computational Methods and Clinical Applications for Spine Imaging (MICCAI CSI), pages 1-13, 2013.
doi bibtex
doi bibtex
@INPROCEEDINGS{miccai_csi2013,
OPTADDRESS = {},
AUTHOR = {Jeremy Kawahara and Chris McIntosh and Roger Tam and
Ghassan Hamarneh},
BOOKTITLE = {Medical Image Computing and Computer-Assisted Intervention Workshop on Computational Methods and Clinical Applications for Spine Imaging (MICCAI CSI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {1-13},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Novel Morphological and Appearance Features for
Predicting Physical Disability from MR Images in Multiple Sclerosis
Patients},
OPTVOLUME = {},
YEAR = {2013},
OPTABSTRACT = {},
DOI = {10.1007/978-3-319-07269-2_6},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Shape Modelling and Analysis},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai_csi2013.pdf}
}Cloud computing for fast prediction of chemical activity. Cała, J., Hiden, H., Woodman, S., & Watson, P. Future Generation Computer Systems, 29(7):1860-1869, 2, 2013.
Paper Website abstract bibtex
Paper Website abstract bibtex Quantitative Structure-Activity Relationships (QSAR) is a method for creating models that can predict certain properties of compounds. It is of growing importance in the design of new drugs. The quantity of data now available for building models is increasing rapidly, which has the advantage that more accurate models can be created, for a wider range of properties. However the disadvantage is that the amount of computation required for model building has also dramatically increased. Therefore, it became vital to find a way to accelerate this process. We have achieved this by exploiting parallelism in searching the QSAR model space for the best models. This paper shows how the cloud computing paradigm can be a good fit to this approach. It describes the design and implementation of a tool for exploring the model space that exploits our e-Science Central cloud platform. We report on the scalability achieved and the experiences gained when designing the solution. The acceleration and absolute performance achieved is much greater than for existing QSAR solutions, creating the potential for new, interesting research, and the exploitation of this approach to accelerate other types of applications.
@article{
title = {Cloud computing for fast prediction of chemical activity},
type = {article},
year = {2013},
identifiers = {[object Object]},
keywords = {Cloud computing,Machine learning,Performance evaluation,Quantitative Structure-Activity Relationships,Scalability},
pages = {1860-1869},
volume = {29},
websites = {http://www.sciencedirect.com/science/article/pii/S0167739X13000253,http://dx.doi.org/10.1016/j.future.2013.01.011,http://linkinghub.elsevier.com/retrieve/pii/S0167739X13000253},
month = {2},
id = {f79ae7a6-d6fa-32bc-85d7-47f91aa10212},
created = {2015-04-27T14:18:27.000Z},
accessed = {2013-02-11},
file_attached = {true},
profile_id = {7e3372c7-5f90-383e-b69f-965cfb3d10ea},
last_modified = {2017-03-10T08:58:51.808Z},
read = {true},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
citation_key = {Cala2013},
folder_uuids = {2556a258-5ac8-427e-8a49-a9d886b3f38a},
private_publication = {false},
abstract = {Quantitative Structure-Activity Relationships (QSAR) is a method for creating models that can predict certain properties of compounds. It is of growing importance in the design of new drugs. The quantity of data now available for building models is increasing rapidly, which has the advantage that more accurate models can be created, for a wider range of properties. However the disadvantage is that the amount of computation required for model building has also dramatically increased. Therefore, it became vital to find a way to accelerate this process. We have achieved this by exploiting parallelism in searching the QSAR model space for the best models. This paper shows how the cloud computing paradigm can be a good fit to this approach. It describes the design and implementation of a tool for exploring the model space that exploits our e-Science Central cloud platform. We report on the scalability achieved and the experiences gained when designing the solution. The acceleration and absolute performance achieved is much greater than for existing QSAR solutions, creating the potential for new, interesting research, and the exploitation of this approach to accelerate other types of applications.},
bibtype = {article},
author = {Cała, Jacek and Hiden, Hugo and Woodman, Simon and Watson, Paul},
journal = {Future Generation Computer Systems},
number = {7}
}2012 (7)
Locally-Adaptive Similarity Metric for Deformable Medical Image Registration. Tang, L. Y. W., Hero, A. O, & Hamarneh, G. In IEEE International Symposium on Biomedical Imaging (IEEE ISBI), pages 728-731, 2012.
doi bibtex
doi bibtex
@INPROCEEDINGS{isbi2012a,
OPTADDRESS = {},
AUTHOR = {Lisa Y. W. Tang and Alfred O Hero and Ghassan Hamarneh},
BOOKTITLE = {IEEE International Symposium on Biomedical Imaging (IEEE ISBI)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {728-731},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Locally-Adaptive Similarity Metric for Deformable
Medical Image Registration},
OPTVOLUME = {},
YEAR = {2012},
OPTABSTRACT = {},
DOI = {10.1109/ISBI.2012.6235651},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Machine Learning, Registration and Matching},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/isbi2012a.pdf}
}Transfer learning for cross-company software defect prediction. Ma, Y., Luo, G., Zeng, X., & Chen, A. Information and Software Technology, 54(3):248--256, 2012.
Paper doi abstract bibtex
Paper doi abstract bibtex Context Software defect prediction studies usually built models using within-company data, but very few focused on the prediction models trained with cross-company data. It is difficult to employ these models which are built on the within-company data in practice, because of the lack of these local data repositories. Recently, transfer learning has attracted more and more attention for building classifier in target domain using the data from related source domain. It is very useful in cases when distributions of training and test instances differ, but is it appropriate for cross-company software defect prediction? Objective In this paper, we consider the cross-company defect prediction scenario where source and target data are drawn from different companies. In order to harness cross company data, we try to exploit the transfer learning method to build faster and highly effective prediction model. Method Unlike the prior works selecting training data which are similar from the test data, we proposed a novel algorithm called Transfer Naive Bayes (TNB), by using the information of all the proper features in training data. Our solution estimates the distribution of the test data, and transfers cross-company data information into the weights of the training data. On these weighted data, the defect prediction model is built. Results This article presents a theoretical analysis for the comparative methods, and shows the experiment results on the data sets from different organizations. It indicates that TNB is more accurate in terms of AUC (The area under the receiver operating characteristic curve), within less runtime than the state of the art methods. Conclusion It is concluded that when there are too few local training data to train good classifiers, the useful knowledge from different-distribution training data on feature level may help. We are optimistic that our transfer learning method can guide optimal resource allocation strategies, which may reduce software testing cost and increase effectiveness of software testing process.
@article{ ma_transfer_2012,
title = {Transfer learning for cross-company software defect prediction},
volume = {54},
issn = {0950-5849},
url = {http://www.sciencedirect.com/science/article/pii/S0950584911001996},
doi = {10.1016/j.infsof.2011.09.007},
abstract = {Context
Software defect prediction studies usually built models using within-company data, but very few focused on the prediction models trained with cross-company data. It is difficult to employ these models which are built on the within-company data in practice, because of the lack of these local data repositories. Recently, transfer learning has attracted more and more attention for building classifier in target domain using the data from related source domain. It is very useful in cases when distributions of training and test instances differ, but is it appropriate for cross-company software defect prediction?
Objective
In this paper, we consider the cross-company defect prediction scenario where source and target data are drawn from different companies. In order to harness cross company data, we try to exploit the transfer learning method to build faster and highly effective prediction model.
Method
Unlike the prior works selecting training data which are similar from the test data, we proposed a novel algorithm called Transfer Naive Bayes (TNB), by using the information of all the proper features in training data. Our solution estimates the distribution of the test data, and transfers cross-company data information into the weights of the training data. On these weighted data, the defect prediction model is built.
Results
This article presents a theoretical analysis for the comparative methods, and shows the experiment results on the data sets from different organizations. It indicates that TNB is more accurate in terms of AUC (The area under the receiver operating characteristic curve), within less runtime than the state of the art methods.
Conclusion
It is concluded that when there are too few local training data to train good classifiers, the useful knowledge from different-distribution training data on feature level may help. We are optimistic that our transfer learning method can guide optimal resource allocation strategies, which may reduce software testing cost and increase effectiveness of software testing process.},
number = {3},
urldate = {2014-07-21TZ},
journal = {Information and Software Technology},
author = {Ma, Ying and Luo, Guangchun and Zeng, Xue and Chen, Aiguo},
year = {2012},
keywords = {Different distribution, Machine learning, Naive Bayes, Software defect prediction, Transfer learning, _done},
pages = {248--256}
}Triaging incoming change requests: Bug or commit history, or code authorship?. Linares-Vásquez, M., Hossen, K., Dang, H., Kagdi, H., Gethers, M., & Poshyvanyk, D. In Software Maintenance (ICSM), 2012 28th IEEE International Conference on, pages 451–460, September, 2012.
doi bibtex
doi bibtex
@inproceedings{linares-vasquez_triaging_2012,
title = {Triaging incoming change requests: {Bug} or commit history, or code authorship?},
doi = {10.1109/ICSM.2012.6405306},
booktitle = {Software {Maintenance} ({ICSM}), 2012 28th {IEEE} {International} {Conference} on},
author = {Linares-Vásquez, M. and Hossen, K. and Dang, Hoang and Kagdi, H. and Gethers, M. and Poshyvanyk, D.},
month = sep,
year = {2012},
keywords = {Accuracy, ArgoUML, Data mining, JEdit, Large scale integration, MuCommander, Software maintenance, Support vector machines, Unified modeling language, bug fixing, change request, code authorship, commit log history, expert developer recommendations, header comments, information retrieval, information retrieval technique, learning (artificial intelligence), machine learning, open source projects, program debugging, public domain software, recommendation accuracies, recommender systems, software change request textual description, source code authorship information, source code files, triaging},
pages = {451--460},
}Learning and Inference in Probabilistic Classifier Chains with Beam Search. Kumar, A., Vembu, S., Menon, A. K., & Elkan, C. In Machine Learning and Knowledge Discovery in Databases, pages 665--680. Springer Berlin Heidelberg, 2012.
Paper doi bibtex
Paper doi bibtex @incollection{Kumar_2012,
Author = {Abhishek Kumar and Shankar Vembu and Aditya Krishna Menon and Charles Elkan},
Booktitle = {Machine Learning and Knowledge Discovery in Databases},
Date-Modified = {2017-09-27 09:02:25 +0000},
Doi = {10.1007/978-3-642-33460-3_48},
Keywords = {machine learning},
Pages = {665--680},
Publisher = {Springer Berlin Heidelberg},
Title = {Learning and Inference in Probabilistic Classifier Chains with Beam Search},
Url = {https://doi.org/10.1007%2F978-3-642-33460-3_48},
Year = 2012,
Bdsk-Url-1 = {https://doi.org/10.1007%2F978-3-642-33460-3_48},
Bdsk-Url-2 = {http://dx.doi.org/10.1007/978-3-642-33460-3_48}}Learning Features for Streak Detection in Dermoscopic Color Images using Localized Radial Flux of Principal Intensity Curvature. Mirzaalian, H., Lee, T., & Hamarneh, G. In IEEE workshop on Mathematical Methods for Biomedical Image Analysis (IEEE MMBIA), pages 97-101, 2012.
doi bibtex
doi bibtex
@INPROCEEDINGS{mmbia2012d,
OPTADDRESS = {},
AUTHOR = {Hengameh Mirzaalian and Tim Lee and Ghassan Hamarneh},
BOOKTITLE = {IEEE workshop on Mathematical Methods for Biomedical Image Analysis (IEEE MMBIA)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {97-101},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Learning Features for Streak Detection in Dermoscopic
Color Images using Localized Radial Flux of Principal Intensity
Curvature},
OPTVOLUME = {},
YEAR = {2012},
OPTABSTRACT = {},
DOI = {10.1109/MMBIA.2012.6164758},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Color, Machine Learning,
Anatomical Trees and Tubular Structures, Dermatology},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/mmbia2012d.pdf}
}Intelligent system for predicting wireless sensor network performance in on-demand deployments. Otero, C., Kostanic, I., Peter, A., Ejnioui, A., & Daniel Otero, L. In 2012 IEEE Conference on Open Systems, ICOS 2012, 2012.
abstract bibtex
abstract bibtex
The need for advanced tools that provide efficient design and planning of on-demand deployment of wireless sensor networks (WSN) is critical for meeting our nation's demand for increased intelligence, reconnaissance, and surveillance in numerous safety-critical applications. For practical applications, WSN deployments can be time-consuming and error-prone, since they have the utmost challenge of guaranteeing connectivity and proper area coverage upon deployment. This creates an unmet demand for decision-support systems that help manage this complex process. This paper presents research-in-progress to develop an advanced decision-support system for predicting the optimal deployment of wireless sensor nodes within an area of interest. The proposed research will have significant impact on the future application of WSN technology, specifically in the emergency response, environmental quality, national security, and engineering education domains. © 2012 IEEE.
@inProceedings{
title = {Intelligent system for predicting wireless sensor network performance in on-demand deployments},
type = {inProceedings},
year = {2012},
identifiers = {[object Object]},
keywords = {[deployments, image processing, machine learning,},
id = {0ffb2f80-1143-3e12-aefa-27bdf5d9583d},
created = {2017-09-15T03:01:29.590Z},
file_attached = {false},
profile_id = {2d070a75-9633-3c98-946c-f010aa829da1},
last_modified = {2017-09-15T03:02:20.314Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {false},
hidden = {false},
folder_uuids = {b8567312-1bd9-42f5-921f-ade5c8cc8aba},
private_publication = {false},
abstract = {The need for advanced tools that provide efficient design and planning of on-demand deployment of wireless sensor networks (WSN) is critical for meeting our nation's demand for increased intelligence, reconnaissance, and surveillance in numerous safety-critical applications. For practical applications, WSN deployments can be time-consuming and error-prone, since they have the utmost challenge of guaranteeing connectivity and proper area coverage upon deployment. This creates an unmet demand for decision-support systems that help manage this complex process. This paper presents research-in-progress to develop an advanced decision-support system for predicting the optimal deployment of wireless sensor nodes within an area of interest. The proposed research will have significant impact on the future application of WSN technology, specifically in the emergency response, environmental quality, national security, and engineering education domains. © 2012 IEEE.},
bibtype = {inProceedings},
author = {Otero, C.E. and Kostanic, I. and Peter, A. and Ejnioui, A. and Daniel Otero, L.},
booktitle = {2012 IEEE Conference on Open Systems, ICOS 2012}
}2011 (5)
Response prediction using collaborative filtering with hierarchies and side-information. Menon, A. K., Chitrapura, K., Garg, S., Agarwal, D., & Kota, N. In Proceedings of the 17th ACM SIGKDD international conference on Knowledge discovery and data mining - KDD \textquotesingle11, 2011. ACM Press.
Paper doi bibtex
Paper doi bibtex @inproceedings{Menon_2011b,
Author = {Aditya Krishna Menon and Krishna-Prasad Chitrapura and Sachin Garg and Deepak Agarwal and Nagaraj Kota},
Booktitle = {Proceedings of the 17th {ACM} {SIGKDD} international conference on Knowledge discovery and data mining - {KDD} {\textquotesingle}11},
Date-Modified = {2017-09-27 08:53:52 +0000},
Doi = {10.1145/2020408.2020436},
Keywords = {machine learning},
Publisher = {{ACM} Press},
Title = {Response prediction using collaborative filtering with hierarchies and side-information},
Url = {https://doi.org/10.1145%2F2020408.2020436},
Year = 2011,
Bdsk-Url-1 = {https://doi.org/10.1145%2F2020408.2020436},
Bdsk-Url-2 = {http://dx.doi.org/10.1145/2020408.2020436}}Fast Algorithms for Approximating the Singular Value Decomposition. Menon, A. K. & Elkan, C. ACM Transactions on Knowledge Discovery from Data, 5(2):1--36, Association for Computing Machinery (ACM), feb, 2011.
Paper doi bibtex
Paper doi bibtex @article{Menon_2011,
Author = {Aditya Krishna Menon and Charles Elkan},
Date-Modified = {2017-09-27 08:53:52 +0000},
Doi = {10.1145/1921632.1921639},
Journal = {{ACM} Transactions on Knowledge Discovery from Data},
Keywords = {machine learning},
Month = {feb},
Number = {2},
Pages = {1--36},
Publisher = {Association for Computing Machinery ({ACM})},
Title = {Fast Algorithms for Approximating the Singular Value Decomposition},
Url = {https://doi.org/10.1145%2F1921632.1921639},
Volume = {5},
Year = 2011,
Bdsk-Url-1 = {https://doi.org/10.1145%2F1921632.1921639},
Bdsk-Url-2 = {http://dx.doi.org/10.1145/1921632.1921639}}Link Prediction via Matrix Factorization. Menon, A. K. & Elkan, C. In Machine Learning and Knowledge Discovery in Databases, pages 437--452. Springer Berlin Heidelberg, 2011.
Paper doi bibtex
Paper doi bibtex @incollection{Menon_2011c,
Author = {Aditya Krishna Menon and Charles Elkan},
Booktitle = {Machine Learning and Knowledge Discovery in Databases},
Date-Modified = {2017-09-27 08:53:52 +0000},
Doi = {10.1007/978-3-642-23783-6_28},
Keywords = {machine learning},
Pages = {437--452},
Publisher = {Springer Berlin Heidelberg},
Title = {Link Prediction via Matrix Factorization},
Url = {https://doi.org/10.1007%2F978-3-642-23783-6_28},
Year = 2011,
Bdsk-Url-1 = {https://doi.org/10.1007%2F978-3-642-23783-6_28},
Bdsk-Url-2 = {http://dx.doi.org/10.1007/978-3-642-23783-6_28}}3D shape analysis of thigh muscles: people with Chronic Obstructive Pulmonary Disease versus healthy older adults. HajGhanbari, B., Hamarneh, G., Changizi, N., Ward, A., & Reid, W. D. In Canadian Physiotherapy Association Congress, pages 1, 2011.
bibtex
bibtex
@INPROCEEDINGS{cpa2011,
OPTADDRESS = {},
AUTHOR = {Bahareh HajGhanbari and Ghassan Hamarneh and
Neda Changizi and Aaron Ward and W. Darlene Reid},
BOOKTITLE = {Canadian Physiotherapy Association Congress},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {1},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {3D shape analysis of thigh muscles: people with Chronic
Obstructive Pulmonary Disease versus healthy older adults},
OPTVOLUME = {},
YEAR = {2011},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Shape Modelling and Analysis, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/cpa2011.pdf}
}Combining functional networks and sensitivity analysis as wrapper method for feature selection. Sánchez-Maroño, N. & Alonso-Betanzos, A. Expert Systems with Applications, 38(10):12930-12938, 2011.
doi bibtex
doi bibtex
@article{ Sanchez2011,
author = {Noelia Sánchez-Maroño and Amparo Alonso-Betanzos},
title = {{C}ombining functional networks and sensitivity analysis as wrapper method for feature selection},
journal = {Expert Systems with Applications},
year = {2011},
volume = {38},
number = {10},
pages = {12930-12938},
doi = {10.1016/j.eswa.2011.04.089},
keywords = {feature selection, wrapper methods, machine learning, neural networks}
}2010 (1)
A Log-Linear Model with Latent Features for Dyadic Prediction. Menon, A. K. & Elkan, C. In 2010 IEEE International Conference on Data Mining, dec, 2010. IEEE.
Paper doi bibtex
Paper doi bibtex @inproceedings{Menon_2010,
Author = {Aditya Krishna Menon and Charles Elkan},
Booktitle = {2010 {IEEE} International Conference on Data Mining},
Date-Modified = {2017-09-27 08:54:22 +0000},
Doi = {10.1109/icdm.2010.148},
Keywords = {machine learning},
Month = {dec},
Publisher = {{IEEE}},
Title = {A Log-Linear Model with Latent Features for Dyadic Prediction},
Url = {https://doi.org/10.1109%2Ficdm.2010.148},
Year = 2010,
Bdsk-Url-1 = {https://doi.org/10.1109%2Ficdm.2010.148},
Bdsk-Url-2 = {http://dx.doi.org/10.1109/icdm.2010.148}}2009 (2)
ViridiScope: Design and Implementation of a Fine Grained Power Monitoring System for Homes. Kim, Y., Schmid, T., Charbiwala, Z. M., & Srivastava, M. B. In Proceedings of the 11th International Conference on Ubiquitous Computing, of Ubicomp '09, pages 245--254, New York, NY, USA, 2009. ACM.
Paper doi abstract bibtex
Paper doi abstract bibtex A key prerequisite for residential energy conservation is knowing when and where energy is being spent. Unfortunately, the current generation of energy reporting devices only provide partial and coarse grained information or require expensive professional installation. This limitation stems from the presumption that calculating per-appliance consumption requires per-appliance current measurements. However, since appliances typically emit measurable signals when they are consuming energy, we can estimate their consumption using indirect sensing. This paper presents ViridiScope, a fine-grained power monitoring system that furnishes users with an economical, self-calibrating tool that provides power consumption of virtually every appliance in the home. ViridiScope uses ambient signals from inexpensive sensors placed near appliances to estimate power consumption, thus no in-line sensor is necessary. We use a model-based machine learning algorithm that automates the sensor calibration process. Through experiments in a real house, we show that ViridiScope can estimate the end-point power consumption within 10% error.
@inproceedings{ kim_viridiscope:_2009,
address = {New York, {NY}, {USA}},
series = {Ubicomp '09},
title = {{ViridiScope}: Design and Implementation of a Fine Grained Power Monitoring System for Homes},
isbn = {978-1-60558-431-7},
shorttitle = {{ViridiScope}},
url = {http://doi.acm.org/10.1145/1620545.1620582},
doi = {10.1145/1620545.1620582},
abstract = {A key prerequisite for residential energy conservation is knowing when and where energy is being spent. Unfortunately, the current generation of energy reporting devices only provide partial and coarse grained information or require expensive professional installation. This limitation stems from the presumption that calculating per-appliance consumption requires per-appliance current measurements. However, since appliances typically emit measurable signals when they are consuming energy, we can estimate their consumption using indirect sensing. This paper presents {ViridiScope}, a fine-grained power monitoring system that furnishes users with an economical, self-calibrating tool that provides power consumption of virtually every appliance in the home. {ViridiScope} uses ambient signals from inexpensive sensors placed near appliances to estimate power consumption, thus no in-line sensor is necessary. We use a model-based machine learning algorithm that automates the sensor calibration process. Through experiments in a real house, we show that {ViridiScope} can estimate the end-point power consumption within 10% error.},
urldate = {2014-09-04TZ},
booktitle = {Proceedings of the 11th International Conference on Ubiquitous Computing},
publisher = {{ACM}},
author = {Kim, Younghun and Schmid, Thomas and Charbiwala, Zainul M. and Srivastava, Mani B.},
year = {2009},
keywords = {adaptive sensor calibration, machine learning, nonintrusive and spatially distributed sensing},
pages = {245--254}
}Reforestation planning using Bayesian networks. Ordóñez Galán, C., Matías, J., Rivas, T., & Bastante, F. Environmental Modelling & Software, 24(11):1285-1292, 11, 2009.
Website doi abstract bibtex
Website doi abstract bibtex The aim of this research was to construct a reforestation model for woodland located in the basin of the river Liébana (NW Spain). This is essentially a pattern recognition problem: the class labels are types of woodland, and the variables for each point are environmental coordinates (referring to altitude, slope, rainfall, lithology, etc.). The model trained using data for existing wooded areas will serve as a guideline for the reforestation of deforested areas. Nonetheless, with a view to tackling reforestation from a more informed perspective, of interest is an interpretable model of relationships existing not just between woodland type and environmental variables but also between and among the environmental variables themselves. For this reason we used Bayesian networks, as a tool that is capable of constructing a causal model of the relationships existing between all the variables represented in the model. The prediction results obtained were compared with those for classical linear techniques, neural networks and support vector machines.
@article{
title = {Reforestation planning using Bayesian networks},
type = {article},
year = {2009},
keywords = {Bayesian networks,Environmental variables,GIS,Machine learning,Reforestation},
pages = {1285-1292},
volume = {24},
websites = {http://www.sciencedirect.com/science/article/pii/S1364815209001224},
month = {11},
id = {e3ab2c9a-49e3-3ad0-8389-7ef8721d92ad},
created = {2015-04-11T19:52:26.000Z},
accessed = {2015-04-11},
file_attached = {false},
profile_id = {95e10851-cdf3-31de-9f82-1ab629e601b0},
group_id = {71a29c65-85d2-3809-a3a1-fe4a94dc78d2},
last_modified = {2017-03-14T14:27:45.955Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {true},
hidden = {false},
private_publication = {false},
abstract = {The aim of this research was to construct a reforestation model for woodland located in the basin of the river Liébana (NW Spain). This is essentially a pattern recognition problem: the class labels are types of woodland, and the variables for each point are environmental coordinates (referring to altitude, slope, rainfall, lithology, etc.). The model trained using data for existing wooded areas will serve as a guideline for the reforestation of deforested areas. Nonetheless, with a view to tackling reforestation from a more informed perspective, of interest is an interpretable model of relationships existing not just between woodland type and environmental variables but also between and among the environmental variables themselves. For this reason we used Bayesian networks, as a tool that is capable of constructing a causal model of the relationships existing between all the variables represented in the model. The prediction results obtained were compared with those for classical linear techniques, neural networks and support vector machines.},
bibtype = {article},
author = {Ordóñez Galán, C. and Matías, J.M. and Rivas, T. and Bastante, F.G.},
doi = {10.1016/j.envsoft.2009.05.009},
journal = {Environmental Modelling & Software},
number = {11}
}2008 (1)
3D Bicipital Groove Shape Analysis and Relationship to Tendopathy. Ward, A., Hamarneh, G., & Schweitzer, M. Journal of Digital Imaging, 21(2):219-234, 2008.
doi bibtex
doi bibtex
@ARTICLE{jdi2008a,
AUTHOR = {Aaron Ward and Ghassan Hamarneh and Mark Schweitzer},
JOURNAL = {Journal of Digital Imaging},
OPTMONTH = {},
OPTNOTE = {},
NUMBER = {2},
PAGES = {219-234},
TITLE = {3D Bicipital Groove Shape Analysis and Relationship to
Tendopathy},
VOLUME = {21},
YEAR = {2008},
OPTABSTRACT = {},
DOI = {10.1007/s10278-007-9027-6},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Shape Modelling and Analysis, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/jdi2008a.pdf}
}2007 (2)
Anatomical Shape Analysis: Exploring the Relationship between Shape and Pathology. Ward, A., Hamarneh, G., & Schweitzer, M. In CIHR National Research Poster Competition, Canadian Student Health Research Forum (CSHRF), Winnipeg, June 6-7, 2007.
bibtex
bibtex
@INPROCEEDINGS{cshrf2007,
OPTADDRESS = {},
AUTHOR = {Aaron Ward and Ghassan Hamarneh and Mark Schweitzer},
BOOKTITLE = {CIHR National Research Poster Competition, Canadian Student Health Research Forum (CSHRF), Winnipeg, June 6-7},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
OPTPAGES = {},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {Anatomical Shape Analysis: Exploring the Relationship
between Shape and Pathology},
OPTVOLUME = {},
YEAR = {2007},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Shape Modelling and Analysis, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/cshrf2007.pdf}
}An overview of anomaly detection techniques: Existing solutions and latest technological trends. Patcha, A. & Park, J. Computer Networks, 51(12):3448-3470, 8, 2007.
Paper abstract bibtex
Paper abstract bibtex As advances in networking technology help to connect the distant corners of the globe and as the Internet continues to expand its influence as a medium for communications and commerce, the threat from spammers, attackers and criminal enterprises has also grown accordingly. It is the prevalence of such threats that has made intrusion detection systems—the cyberspace’s equivalent to the burglar alarm—join ranks with firewalls as one of the fundamental technologies for network security. However, today’s commercially available intrusion detection systems are predominantly signature-based intrusion detection systems that are designed to detect known attacks by utilizing the signatures of those attacks. Such systems require frequent rule-base updates and signature updates, and are not capable of detecting unknown attacks. In contrast, anomaly detection systems, a subset of intrusion detection systems, model the normal system/network behavior which enables them to be extremely effective in finding and foiling both known as well as unknown or “zero day” attacks. While anomaly detection systems are attractive conceptually, a host of technological problems need to be overcome before they can be widely adopted. These problems include: high false alarm rate, failure to scale to gigabit speeds, etc. In this paper, we provide a comprehensive survey of anomaly detection systems and hybrid intrusion detection systems of the recent past and present. We also discuss recent technological trends in anomaly detection and identify open problems and challenges in this area.
@article{
id = {8d92f507-1f5e-3a12-a412-07cf9d45e770},
title = {An overview of anomaly detection techniques: Existing solutions and latest technological trends},
type = {article},
year = {2007},
identifiers = {[object Object]},
keywords = {Anomaly detection,Data mining,Machine learning,Statistical anomaly detection,Survey},
created = {2015-04-08T01:47:24.000Z},
pages = {3448-3470},
volume = {51},
websites = {http://www.sciencedirect.com/science/article/pii/S138912860700062X},
month = {8},
accessed = {2014-07-10},
file_attached = {true},
profile_id = {595d3c8c-e6f7-3225-b677-525603e28b20},
last_modified = {2015-04-16T16:39:23.000Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
abstract = {As advances in networking technology help to connect the distant corners of the globe and as the Internet continues to expand its influence as a medium for communications and commerce, the threat from spammers, attackers and criminal enterprises has also grown accordingly. It is the prevalence of such threats that has made intrusion detection systems—the cyberspace’s equivalent to the burglar alarm—join ranks with firewalls as one of the fundamental technologies for network security. However, today’s commercially available intrusion detection systems are predominantly signature-based intrusion detection systems that are designed to detect known attacks by utilizing the signatures of those attacks. Such systems require frequent rule-base updates and signature updates, and are not capable of detecting unknown attacks. In contrast, anomaly detection systems, a subset of intrusion detection systems, model the normal system/network behavior which enables them to be extremely effective in finding and foiling both known as well as unknown or “zero day” attacks. While anomaly detection systems are attractive conceptually, a host of technological problems need to be overcome before they can be widely adopted. These problems include: high false alarm rate, failure to scale to gigabit speeds, etc. In this paper, we provide a comprehensive survey of anomaly detection systems and hybrid intrusion detection systems of the recent past and present. We also discuss recent technological trends in anomaly detection and identify open problems and challenges in this area.},
bibtype = {article},
author = {Patcha, Animesh and Park, Jung-Min},
journal = {Computer Networks},
number = {12}
}2006 (2)
3D Shape Analysis of the Supraspinatus Muscle. Ward, A., Hamarneh, G., Ashry, R., & Schweitzer, M. In Medical Image Computing and Computer-Assisted Intervention Joint Diseases Workshop (MICCAI JD), pages 96-103, 2006.
bibtex
bibtex
@INPROCEEDINGS{miccai_jd2006b,
OPTADDRESS = {},
AUTHOR = {Aaron Ward and Ghassan Hamarneh and Reem Ashry and
Mark Schweitzer},
BOOKTITLE = {Medical Image Computing and Computer-Assisted Intervention Joint Diseases Workshop (MICCAI JD)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {96-103},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {3D Shape Analysis of the Supraspinatus Muscle},
OPTVOLUME = {},
YEAR = {2006},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Shape Modelling and Analysis, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai_jd2006b.pdf}
}3D Shape Description of the Bicipital Groove: Correlation to Pathology. Ward, A., Hamarneh, G., & Schweitzer, M. In Medical Image Computing and Computer-Assisted Intervention Joint Diseases Workshop (MICCAI JD), pages 80-87, 2006.
bibtex
bibtex
@INPROCEEDINGS{miccai_jd2006a,
OPTADDRESS = {},
AUTHOR = {Aaron Ward and Ghassan Hamarneh and Mark Schweitzer},
BOOKTITLE = {Medical Image Computing and Computer-Assisted Intervention Joint Diseases Workshop (MICCAI JD)},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {80-87},
OPTPUBLISHER = {},
OPTSERIES = {},
TITLE = {3D Shape Description of the Bicipital Groove:
Correlation to Pathology},
OPTVOLUME = {},
YEAR = {2006},
OPTABSTRACT = {},
OPTDOI = {},
OPTISBN = {},
OPTISSN = {},
KEYWORDS = {Shape Modelling and Analysis, Machine Learning},
OPTURL = {},
OPTURL-PUBLISHER = {},
PDF = {http://www.cs.sfu.ca/~hamarneh/ecopy/miccai_jd2006a.pdf}
}2005 (1)
A Tutorial on the Cross-Entropy Method. Boer, P., D., E. 2005.
Paper bibtex
Paper bibtex @article{
title = {A Tutorial on the Cross-Entropy Method},
type = {article},
year = {2005},
keywords = {cross-entropy method,machine learning,monte-carlo simulation,randomized optimization},
pages = {19-67},
id = {6231f5d6-0efc-3497-8415-8abb35bac064},
created = {2016-10-17T18:07:57.000Z},
file_attached = {true},
profile_id = {53c9663a-19e4-3f1b-8334-2677a68201f7},
group_id = {0774430b-2ad2-3ef0-ad04-965ed37210f8},
last_modified = {2016-10-17T18:08:18.000Z},
read = {false},
starred = {false},
authored = {false},
confirmed = {false},
hidden = {false},
bibtype = {article},
author = {Boer, Pieter-tjerk D E}
}2004 (1)
A Combined Transmembrane Topology and Signal Peptide Prediction Method. Käll, L., Krogh, A., & Sonnhammer, E. L. L Journal of Molecular Biology, 338(5):1027–1036, May, 2004.
Paper doi abstract bibtex
Paper doi abstract bibtex An inherent problem in transmembrane protein topology prediction and signal peptide prediction is the high similarity between the hydrophobic regions of a transmembrane helix and that of a signal peptide, leading to cross-reaction between the two types of predictions. To improve predictions further, it is therefore important to make a predictor that aims to discriminate between the two classes. In addition, topology information can be gained when successfully predicting a signal peptide leading a transmembrane protein since it dictates that the N terminus of the mature protein must be on the non-cytoplasmic side of the membrane. Here, we present Phobius, a combined transmembrane protein topology and signal peptide predictor. The predictor is based on a hidden Markov model (HMM) that models the different sequence regions of a signal peptide and the different regions of a transmembrane protein in a series of interconnected states. Training was done on a newly assembled and curated dataset. Compared to TMHMM and SignalP, errors coming from cross-prediction between transmembrane segments and signal peptides were reduced substantially by Phobius. False classifications of signal peptides were reduced from 26.1% to 3.9% and false classifications of transmembrane helices were reduced from 19.0% to 7.7%. Phobius was applied to the proteomes of Homo sapiens and Escherichia coli. Here we also noted a drastic reduction of false classifications compared to TMHMM/SignalP, suggesting that Phobius is well suited for whole-genome annotation of signal peptides and transmembrane regions. The method is available at http://phobius.cgb.ki.se/ as well as at http://phobius.binf.ku.dk/
@article{kall_combined_2004,
title = {A {Combined} {Transmembrane} {Topology} and {Signal} {Peptide} {Prediction} {Method}},
volume = {338},
issn = {0022-2836},
url = {http://www.sciencedirect.com/science/article/pii/S0022283604002943},
doi = {10.1016/j.jmb.2004.03.016},
abstract = {An inherent problem in transmembrane protein topology prediction and signal peptide prediction is the high similarity between the hydrophobic regions of a transmembrane helix and that of a signal peptide, leading to cross-reaction between the two types of predictions. To improve predictions further, it is therefore important to make a predictor that aims to discriminate between the two classes. In addition, topology information can be gained when successfully predicting a signal peptide leading a transmembrane protein since it dictates that the N terminus of the mature protein must be on the non-cytoplasmic side of the membrane. Here, we present Phobius, a combined transmembrane protein topology and signal peptide predictor. The predictor is based on a hidden Markov model (HMM) that models the different sequence regions of a signal peptide and the different regions of a transmembrane protein in a series of interconnected states. Training was done on a newly assembled and curated dataset. Compared to TMHMM and SignalP, errors coming from cross-prediction between transmembrane segments and signal peptides were reduced substantially by Phobius. False classifications of signal peptides were reduced from 26.1\% to 3.9\% and false classifications of transmembrane helices were reduced from 19.0\% to 7.7\%. Phobius was applied to the proteomes of Homo sapiens and Escherichia coli. Here we also noted a drastic reduction of false classifications compared to TMHMM/SignalP, suggesting that Phobius is well suited for whole-genome annotation of signal peptides and transmembrane regions. The method is available at http://phobius.cgb.ki.se/ as well as at http://phobius.binf.ku.dk/},
number = {5},
urldate = {2017-11-06TZ},
journal = {Journal of Molecular Biology},
author = {Käll, Lukas and Krogh, Anders and Sonnhammer, Erik L. L},
month = may,
year = {2004},
keywords = {hidden Markov model, machine learning, signal peptide, topology prediction, transmembrane protein},
pages = {1027--1036}
}2002 (2)
Machine Learning in Automated Text Categorization. Sebastiani, F. ACM Comput Surv, 34(1):1–47, ACM, New York, NY, USA, 2002.
doi abstract bibtex
doi abstract bibtex
The automated categorization (or classification) of texts into predefined categories has witnessed a booming interest in the last 10 years, due to the increased availability of documents in digital form and the ensuing need to organize them. In the research community the dominant approach to this problem is based on machine learning techniques: a general inductive process automatically builds a classifier by learning, from a set of preclassified documents, the characteristics of the categories. The advantages of this approach over the knowledge engineering approach (consisting in the manual definition of a classifier by domain experts) are a very good effectiveness, considerable savings in terms of expert labor power, and straightforward portability to different domains. This survey discusses the main approaches to text categorization that fall within the machine learning paradigm. We will discuss in detail issues pertaining to three different problems, namely, document representation, classifier construction, and classifier evaluation.
@Article{sebastiani02machine,
author = {Sebastiani, Fabrizio},
title = {Machine Learning in Automated Text Categorization},
journal = {ACM Comput Surv},
year = {2002},
volume = {34},
number = {1},
pages = {1--47},
issn = {0360-0300},
abstract = {The automated categorization (or classification) of texts into predefined categories has witnessed a booming interest in the last 10 years, due to the increased availability of documents in digital form and the ensuing need to organize them. In the research community the dominant approach to this problem is based on machine learning techniques: a general inductive process automatically builds a classifier by learning, from a set of preclassified documents, the characteristics of the categories. The advantages of this approach over the knowledge engineering approach (consisting in the manual definition of a classifier by domain experts) are a very good effectiveness, considerable savings in terms of expert labor power, and straightforward portability to different domains. This survey discusses the main approaches to text categorization that fall within the machine learning paradigm. We will discuss in detail issues pertaining to three different problems, namely, document representation, classifier construction, and classifier evaluation.},
acmid = {505283},
address = {New York, NY, USA},
comment = {microaveraged F1 und accuracy},
doi = {10.1145/505282.505283},
issue_date = {March 2002},
keywords = {Machine learning, text categorization, text classification},
numpages = {47},
owner = {Sebastian},
publisher = {ACM},
timestamp = {2017.03.05},
}Research abstract for semantic anomaly detection in dynamic data feeds with incomplete specifications. Raz, O. In Proceedings of the 24rd International Conference on Software Engineering, 2002. ICSE 2002, pages 733--734, May, 2002.
abstract bibtex
abstract bibtex
Everyday software must be dependable enough for its intended use. Because this software is not usually mission-critical, it may be cost-effective to detect improper behavior and notify the user or take remedial action. Detecting improper behavior requires a model of proper behavior. Unfortunately, specifications of everyday software are often incomplete and imprecise. The situation is exacerbated when the software incorporates third-party elements such as commercial-off-the-shelf software components, databases, or dynamic data feeds from online data sources. We want to make the use of dynamic data feeds more dependable. We are specifically interested in semantic problems with these feeds-cases in which the data feed is responsive, it delivers well-formed results, but the results are inconsistent, out of range, incorrect, or otherwise unreasonable. We focus on a particular facet of dependability: availability or readiness for usage, and change the fault model from the traditional "fail-silent" (crash failures) to "semantic". We investigate anomaly detection as a step towards increasing the semantic availability of dynamic data feeds.
@inproceedings{ raz_research_2002,
title = {Research abstract for semantic anomaly detection in dynamic data feeds with incomplete specifications},
abstract = {Everyday software must be dependable enough for its intended use. Because this software is not usually mission-critical, it may be cost-effective to detect improper behavior and notify the user or take remedial action. Detecting improper behavior requires a model of proper behavior. Unfortunately, specifications of everyday software are often incomplete and imprecise. The situation is exacerbated when the software incorporates third-party elements such as commercial-off-the-shelf software components, databases, or dynamic data feeds from online data sources. We want to make the use of dynamic data feeds more dependable. We are specifically interested in semantic problems with these feeds-cases in which the data feed is responsive, it delivers well-formed results, but the results are inconsistent, out of range, incorrect, or otherwise unreasonable. We focus on a particular facet of dependability: availability or readiness for usage, and change the fault model from the traditional "fail-silent" (crash failures) to "semantic". We investigate anomaly detection as a step towards increasing the semantic availability of dynamic data feeds.},
booktitle = {Proceedings of the 24rd {International} {Conference} on {Software} {Engineering}, 2002. {ICSE} 2002},
author = {Raz, O.},
month = {May},
year = {2002},
keywords = {Availability, Computer crashes, Computer science, Data mining, Databases, Face detection, Fault model, Feeds, Machine learning, Permission, Stock markets, Training data, _done, dependability, dynamic data feeds, incomplete specifications, inference mechanisms, learning (artificial intelligence), missing specifications, normal behavior, online data sources, proper behavior, readiness for usage, semantic anomaly detection, semantic problems, useful characteristics},
pages = {733--734}
}1999 (1)
An overview of statistical learning theory. Vapnik, V. N. IEEE Trans Neural Netw, 10(5):988–999, 1999.
doi bibtex
doi bibtex
@Article{vapnik99overview,
author = {V. N. Vapnik},
title = {An overview of statistical learning theory},
journal = {IEEE Trans Neural Netw},
year = {1999},
volume = {10},
number = {5},
pages = {988--999},
optmonth = sep,
issn = {1045-9227},
doi = {10.1109/72.788640},
keywords = {estimation theory, generalisation (artificial intelligence), learning (artificial intelligence), statistical analysis, function estimation, generalization conditions, multidimensional function estimation, statistical learning theory, support vector machines, Algorithm design and analysis, Loss measurement, Machine learning, Multidimensional systems, Pattern recognition, Probability distribution, Risk management, Statistical learning, Support vector machines},
}1994 (2)
On a learnability question associated to neural networks with continuous activations (extended abstract). DasGupta, B., Siegelmann, H. T., & Sontag, E. In COLT '94: Proceedings of the seventh annual conference on Computational learning theory, pages 47–56, New York, NY, USA, 1994. ACM Press.
doi bibtex
doi bibtex
@INPROCEEDINGS{181009,
AUTHOR = {B. DasGupta and H. T. Siegelmann and E.D. Sontag},
BOOKTITLE = {COLT '94: Proceedings of the seventh annual conference on Computational learning theory},
TITLE = {On a learnability question associated to neural networks
with continuous activations (extended abstract)},
YEAR = {1994},
ADDRESS = {New York, NY, USA},
OPTCROSSREF = {},
OPTEDITOR = {},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
OPTORGANIZATION = {},
PAGES = {47--56},
PUBLISHER = {ACM Press},
OPTSERIES = {},
OPTVOLUME = {},
KEYWORDS = {analog computing, neural networks,
computational complexity, machine learning},
DOI = {http://doi.acm.org/10.1145/180139.181009}
}On the Intractability of Loading Neural Networks. DasGupta, B., Siegelmann, H., & Sontag, E. In Roychowdhury, V. P., Y., S. K., & A., O., editors, Theoretical Advances in Neural Computation and Learning, pages 357–389. Kluwer Academic Publishers, 1994.
bibtex
bibtex
@INCOLLECTION{DasGupta-book,
AUTHOR = {B. DasGupta and H.T. Siegelmann and E.D. Sontag},
BOOKTITLE = {Theoretical Advances in Neural Computation and Learning},
PUBLISHER = {Kluwer Academic Publishers},
TITLE = {On the Intractability of Loading Neural Networks},
YEAR = {1994},
OPTADDRESS = {},
OPTCHAPTER = {},
OPTCROSSREF = {},
OPTEDITION = {},
EDITOR = {V. P. Roychowdhury and Siu K. Y. and Orlitsky A.},
OPTMONTH = {},
OPTNOTE = {},
OPTNUMBER = {},
PAGES = {357--389},
OPTSERIES = {},
OPTTYPE = {},
OPTVOLUME = {},
KEYWORDS = {analog computing, neural networks,
computational complexity, machine learning},
PDF = {../../FTPDIR/dasgupta_siegelmann_sontag_intractability_loading_neural_networks_bookchapter_springer1994.pdf}
}1991 (1)
Adaptive case-based reasoning. Callan, J. & Fawcett, T. In Proceedings of the Third DARPA Case-Based Reasoning, pages 179-190, 1991. Morgan Kaufmann.
Website bibtex
Website bibtex @inProceedings{
title = {Adaptive case-based reasoning},
type = {inProceedings},
year = {1991},
keywords = {case-based reasoning,machine learning},
pages = {179-190},
websites = {http://scholar.google.com/scholar?hl=en&btnG=Search&q=intitle:Adaptive+case-based+reasoning#5},
publisher = {Morgan Kaufmann},
city = {Washington, D.C.},
id = {fbe8f99f-5078-3ef5-8c78-df574962d440},
created = {2009-08-18T04:22:20.000Z},
accessed = {2012-06-09},
file_attached = {false},
profile_id = {7edb4dd2-367e-3b20-8e26-a7e1d668fa33},
last_modified = {2017-07-27T21:18:51.927Z},
read = {false},
starred = {false},
authored = {true},
confirmed = {true},
hidden = {false},
citation_key = {cal91d},
source_type = {inproceedings},
private_publication = {false},
bibtype = {inProceedings},
author = {Callan, James and Fawcett, Tom},
booktitle = {Proceedings of the Third DARPA Case-Based Reasoning}
}