Approximate Kernel Matrix Computation on GPUs Forlarge Scale Learning Applications. Hussein, M. E. & Abd-Almageed, W. In Proceedings of the 23rd International Conference on Supercomputing, of ICS '09, pages 511–512, 2009. ACM. event-place: Yorktown Heights, NY, USA![Approximate Kernel Matrix Computation on GPUs Forlarge Scale Learning Applications [link]](https://bibbase.org/img/filetypes/link.svg "link")
Paper doi abstract bibtex Kernel-based learning methods require quadratic space and time complexities to compute the kernel matrix. These complexities limit the applicability of kernel methods to large scale problems with millions of data points. In this paper, we introduce a novel representation of kernel matrices on Graphics Processing Units (GPU). The novel representation exploits the sparseness of the kernel matrix to address the space complexity problem. It also respects the guidelines for memory access on GPUs, which are critical for good performance, to address the time complexity problem. Our representation utilizes the locality preserving properties of space filling curves to obtain a band approximation of the kernel matrix. To prove the validity of the representation, we use Affinity Propagation, an unsupervised clustering algorithm, as an example of kernel methods. Experimental results show a 40x speedup of AP using our representation without degradation in clustering performance.
@inproceedings{hussein_approximate_2009,
location = {New York, {NY}, {USA}},
title = {Approximate Kernel Matrix Computation on {GPUs} Forlarge Scale Learning Applications},
rights = {All rights reserved},
isbn = {978-1-60558-498-0},
url = {http://doi.acm.org/10.1145/1542275.1542355},
doi = {10.1145/1542275.1542355},
series = {{ICS} '09},
abstract = {Kernel-based learning methods require quadratic space and time complexities to compute the kernel matrix. These complexities limit the applicability of kernel methods to large scale problems with millions of data points. In this paper, we introduce a novel representation of kernel matrices on Graphics Processing Units ({GPU}). The novel representation exploits the sparseness of the kernel matrix to address the space complexity problem. It also respects the guidelines for memory access on {GPUs}, which are critical for good performance, to address the time complexity problem. Our representation utilizes the locality preserving properties of space filling curves to obtain a band approximation of the kernel matrix. To prove the validity of the representation, we use Affinity Propagation, an unsupervised clustering algorithm, as an example of kernel methods. Experimental results show a 40x speedup of {AP} using our representation without degradation in clustering performance.},
pages = {511--512},
booktitle = {Proceedings of the 23rd International Conference on Supercomputing},
publisher = {{ACM}},
author = {Hussein, Mohamed E. and Abd-Almageed, Wael},
urldate = {2019-05-01},
year = {2009},
note = {event-place: Yorktown Heights, {NY}, {USA}},
keywords = {affinity propagation, gpu, kernel methods, space filling curves, sparse matrices},
file = {ACM Full Text PDF:C\:\\Users\\Mohamed Hussein\\Zotero\\storage\\C6XCZ2V9\\Hussein and Abd-Almageed - 2009 - Approximate Kernel Matrix Computation on GPUs Forl.pdf:application/pdf}
}
Downloads: 0
{"_id":"MGSK4dkWDnHedW5RQ","bibbaseid":"hussein-abdalmageed-approximatekernelmatrixcomputationongpusforlargescalelearningapplications-2009","author_short":["Hussein, M. E.","Abd-Almageed, W."],"bibdata":{"bibtype":"inproceedings","type":"inproceedings","location":"New York, NY, USA","title":"Approximate Kernel Matrix Computation on GPUs Forlarge Scale Learning Applications","rights":"All rights reserved","isbn":"978-1-60558-498-0","url":"http://doi.acm.org/10.1145/1542275.1542355","doi":"10.1145/1542275.1542355","series":"ICS '09","abstract":"Kernel-based learning methods require quadratic space and time complexities to compute the kernel matrix. These complexities limit the applicability of kernel methods to large scale problems with millions of data points. In this paper, we introduce a novel representation of kernel matrices on Graphics Processing Units (GPU). The novel representation exploits the sparseness of the kernel matrix to address the space complexity problem. It also respects the guidelines for memory access on GPUs, which are critical for good performance, to address the time complexity problem. Our representation utilizes the locality preserving properties of space filling curves to obtain a band approximation of the kernel matrix. To prove the validity of the representation, we use Affinity Propagation, an unsupervised clustering algorithm, as an example of kernel methods. Experimental results show a 40x speedup of AP using our representation without degradation in clustering performance.","pages":"511–512","booktitle":"Proceedings of the 23rd International Conference on Supercomputing","publisher":"ACM","author":[{"propositions":[],"lastnames":["Hussein"],"firstnames":["Mohamed","E."],"suffixes":[]},{"propositions":[],"lastnames":["Abd-Almageed"],"firstnames":["Wael"],"suffixes":[]}],"urldate":"2019-05-01","year":"2009","note":"event-place: Yorktown Heights, NY, USA","keywords":"affinity propagation, gpu, kernel methods, space filling curves, sparse matrices","file":"ACM Full Text PDF:C\\:\\\\Users\\\\Mohamed Hussein\\\\Zotero\\\\storage\\\\C6XCZ2V9\\űssein and Abd-Almageed - 2009 - Approximate Kernel Matrix Computation on GPUs Forl.pdf:application/pdf","bibtex":"@inproceedings{hussein_approximate_2009,\n\tlocation = {New York, {NY}, {USA}},\n\ttitle = {Approximate Kernel Matrix Computation on {GPUs} Forlarge Scale Learning Applications},\n\trights = {All rights reserved},\n\tisbn = {978-1-60558-498-0},\n\turl = {http://doi.acm.org/10.1145/1542275.1542355},\n\tdoi = {10.1145/1542275.1542355},\n\tseries = {{ICS} '09},\n\tabstract = {Kernel-based learning methods require quadratic space and time complexities to compute the kernel matrix. These complexities limit the applicability of kernel methods to large scale problems with millions of data points. In this paper, we introduce a novel representation of kernel matrices on Graphics Processing Units ({GPU}). The novel representation exploits the sparseness of the kernel matrix to address the space complexity problem. It also respects the guidelines for memory access on {GPUs}, which are critical for good performance, to address the time complexity problem. Our representation utilizes the locality preserving properties of space filling curves to obtain a band approximation of the kernel matrix. To prove the validity of the representation, we use Affinity Propagation, an unsupervised clustering algorithm, as an example of kernel methods. Experimental results show a 40x speedup of {AP} using our representation without degradation in clustering performance.},\n\tpages = {511--512},\n\tbooktitle = {Proceedings of the 23rd International Conference on Supercomputing},\n\tpublisher = {{ACM}},\n\tauthor = {Hussein, Mohamed E. and Abd-Almageed, Wael},\n\turldate = {2019-05-01},\n\tyear = {2009},\n\tnote = {event-place: Yorktown Heights, {NY}, {USA}},\n\tkeywords = {affinity propagation, gpu, kernel methods, space filling curves, sparse matrices},\n\tfile = {ACM Full Text PDF:C\\:\\\\Users\\\\Mohamed Hussein\\\\Zotero\\\\storage\\\\C6XCZ2V9\\\\Hussein and Abd-Almageed - 2009 - Approximate Kernel Matrix Computation on GPUs Forl.pdf:application/pdf}\n}\n\n","author_short":["Hussein, M. E.","Abd-Almageed, W."],"bibbaseid":"hussein-abdalmageed-approximatekernelmatrixcomputationongpusforlargescalelearningapplications-2009","role":"author","urls":{"Paper":"http://doi.acm.org/10.1145/1542275.1542355"},"keyword":["affinity propagation","gpu","kernel methods","space filling curves","sparse matrices"],"metadata":{"authorlinks":{}}},"bibtype":"inproceedings","biburl":"https://bibbase.org/f/2bJzYjCLapWTtM86s/mehussein-2023.bib","dataSources":["kYvtZ54PgkXqjbteW","dWqYiMkhjrrw3PpB5","mhdykGczo2jDicE3X","havAjNnaG4BxhYWyb"],"keywords":["affinity propagation","gpu","kernel methods","space filling curves","sparse matrices"],"search_terms":["approximate","kernel","matrix","computation","gpus","forlarge","scale","learning","applications","hussein","abd-almageed"],"title":"Approximate Kernel Matrix Computation on GPUs Forlarge Scale Learning Applications","year":2009}