Gaussian-Dirichlet Random Fields for Inference over High Dimensional Categorical Observations

Gaussian-Dirichlet Random Fields for Inference over High Dimensional Categorical Observations. Soucie, J., E., S., Sosik, H., M., & Girdhar, Y. In 2020 IEEE International Conference on Robotics and Automation (ICRA), pages 2924-2931, 5, 2020. IEEE.

Website doi abstract bibtex 2 downloads

We propose a generative model for the spatio-temporal distribution of high dimensional categorical observa- tions. These are commonly produced by robots equipped with an imaging sensor such as a camera, paired with an image classifier, potentially producing observations over thousands of categories. The proposed approach combines the use of Dirichlet distributions to model sparse co-occurrence relations between the observed categories using a latent variable, and Gaussian processes to model the latent variable’s spatio- temporal distribution. Experiments in this paper show that the resulting model is able to efficiently and accurately approximate the temporal distribution of high dimensional categorical measurements such as taxonomic observations of microscopic organisms in the ocean, even in unobserved (held out) locations, far from other samples. This work’s primary motivation is to enable deployment of informative path planning techniques over high dimensional categorical fields, which until now have been limited to scalar or low dimensional vector observations.

@inproceedings{
 title = {Gaussian-Dirichlet Random Fields for Inference over High Dimensional Categorical Observations},
 type = {inproceedings},
 year = {2020},
 pages = {2924-2931},
 websites = {https://arxiv.org/abs/2003.12120,https://ieeexplore.ieee.org/document/9196713/},
 month = {5},
 publisher = {IEEE},
 id = {3387f3c9-dbb7-3fcd-ab69-39c2bf8dd168},
 created = {2019-12-23T20:46:27.270Z},
 file_attached = {true},
 profile_id = {2331788d-b144-3e67-ab8c-4abd7ab569c5},
 last_modified = {2021-01-08T10:30:35.502Z},
 read = {false},
 starred = {false},
 authored = {true},
 confirmed = {true},
 hidden = {false},
 citation_key = {Soucie2020},
 folder_uuids = {3acfd987-bf37-4360-8703-a8c92a1c098e,bc083a4d-5255-4ca2-aaa8-932c32e96fc0},
 private_publication = {false},
 abstract = {We propose a generative model for the spatio-temporal distribution of high dimensional categorical observa- tions. These are commonly produced by robots equipped with an imaging sensor such as a camera, paired with an image classifier, potentially producing observations over thousands of categories. The proposed approach combines the use of Dirichlet distributions to model sparse co-occurrence relations between the observed categories using a latent variable, and Gaussian processes to model the latent variable’s spatio- temporal distribution. Experiments in this paper show that the resulting model is able to efficiently and accurately approximate the temporal distribution of high dimensional categorical measurements such as taxonomic observations of microscopic organisms in the ocean, even in unobserved (held out) locations, far from other samples. This work’s primary motivation is to enable deployment of informative path planning techniques over high dimensional categorical fields, which until now have been limited to scalar or low dimensional vector observations.},
 bibtype = {inproceedings},
 author = {Soucie, John E. San and Sosik, Heidi M. and Girdhar, Yogesh},
 doi = {10.1109/ICRA40945.2020.9196713},
 booktitle = {2020 IEEE International Conference on Robotics and Automation (ICRA)}
}

Downloads: 2

{"_id":"93mdX9ev6DHoR7sCf","bibbaseid":"soucie-sosik-girdhar-gaussiandirichletrandomfieldsforinferenceoverhighdimensionalcategoricalobservations-2020","authorIDs":["3vXBM4S2vXwK6sYLP","5e590f799bb6d2df0100005f","5e60654d9119f0de01000006","5e63ae9a5e3c57de01000138","8CctKmi6QSdpDZhZQ","AdBsWPGPxtrLTFRva","BpKDeE3QZpf9X3e9i","GKxwPDd8nnZzt2D66","GpYAXdHXD7BwCEbTR","H9Y2sgmzyKqaAALMo","KMWsQiocBDitEciM8","KshkYviTCtMA43wiA","QziboxkhDdwvZFswA","h2Dy7nzBQGgf7Jm7J","n5RKKFHmvyHk9nQoK","nLtoDjh2WdkwX8ou7","r2AdQsYqi9xWr2Xvn","rYZSywniN29pa3ff5","ux58nAkaFoTsYtjM4","wYyZDsF93sT3wy97F","zqzQF9qtL5fqCwsDi"],"author_short":["Soucie, J., E., S.","Sosik, H., M.","Girdhar, Y."],"bibdata":{"title":"Gaussian-Dirichlet Random Fields for Inference over High Dimensional Categorical Observations","type":"inproceedings","year":"2020","pages":"2924-2931","websites":"https://arxiv.org/abs/2003.12120,https://ieeexplore.ieee.org/document/9196713/","month":"5","publisher":"IEEE","id":"3387f3c9-dbb7-3fcd-ab69-39c2bf8dd168","created":"2019-12-23T20:46:27.270Z","file_attached":"true","profile_id":"2331788d-b144-3e67-ab8c-4abd7ab569c5","last_modified":"2021-01-08T10:30:35.502Z","read":false,"starred":false,"authored":"true","confirmed":"true","hidden":false,"citation_key":"Soucie2020","folder_uuids":"3acfd987-bf37-4360-8703-a8c92a1c098e,bc083a4d-5255-4ca2-aaa8-932c32e96fc0","private_publication":false,"abstract":"We propose a generative model for the spatio-temporal distribution of high dimensional categorical observa- tions. These are commonly produced by robots equipped with an imaging sensor such as a camera, paired with an image classifier, potentially producing observations over thousands of categories. The proposed approach combines the use of Dirichlet distributions to model sparse co-occurrence relations between the observed categories using a latent variable, and Gaussian processes to model the latent variable’s spatio- temporal distribution. Experiments in this paper show that the resulting model is able to efficiently and accurately approximate the temporal distribution of high dimensional categorical measurements such as taxonomic observations of microscopic organisms in the ocean, even in unobserved (held out) locations, far from other samples. This work’s primary motivation is to enable deployment of informative path planning techniques over high dimensional categorical fields, which until now have been limited to scalar or low dimensional vector observations.","bibtype":"inproceedings","author":"Soucie, John E. San and Sosik, Heidi M. and Girdhar, Yogesh","doi":"10.1109/ICRA40945.2020.9196713","booktitle":"2020 IEEE International Conference on Robotics and Automation (ICRA)","bibtex":"@inproceedings{\n title = {Gaussian-Dirichlet Random Fields for Inference over High Dimensional Categorical Observations},\n type = {inproceedings},\n year = {2020},\n pages = {2924-2931},\n websites = {https://arxiv.org/abs/2003.12120,https://ieeexplore.ieee.org/document/9196713/},\n month = {5},\n publisher = {IEEE},\n id = {3387f3c9-dbb7-3fcd-ab69-39c2bf8dd168},\n created = {2019-12-23T20:46:27.270Z},\n file_attached = {true},\n profile_id = {2331788d-b144-3e67-ab8c-4abd7ab569c5},\n last_modified = {2021-01-08T10:30:35.502Z},\n read = {false},\n starred = {false},\n authored = {true},\n confirmed = {true},\n hidden = {false},\n citation_key = {Soucie2020},\n folder_uuids = {3acfd987-bf37-4360-8703-a8c92a1c098e,bc083a4d-5255-4ca2-aaa8-932c32e96fc0},\n private_publication = {false},\n abstract = {We propose a generative model for the spatio-temporal distribution of high dimensional categorical observa- tions. These are commonly produced by robots equipped with an imaging sensor such as a camera, paired with an image classifier, potentially producing observations over thousands of categories. The proposed approach combines the use of Dirichlet distributions to model sparse co-occurrence relations between the observed categories using a latent variable, and Gaussian processes to model the latent variable’s spatio- temporal distribution. Experiments in this paper show that the resulting model is able to efficiently and accurately approximate the temporal distribution of high dimensional categorical measurements such as taxonomic observations of microscopic organisms in the ocean, even in unobserved (held out) locations, far from other samples. This work’s primary motivation is to enable deployment of informative path planning techniques over high dimensional categorical fields, which until now have been limited to scalar or low dimensional vector observations.},\n bibtype = {inproceedings},\n author = {Soucie, John E. San and Sosik, Heidi M. and Girdhar, Yogesh},\n doi = {10.1109/ICRA40945.2020.9196713},\n booktitle = {2020 IEEE International Conference on Robotics and Automation (ICRA)}\n}","author_short":["Soucie, J., E., S.","Sosik, H., M.","Girdhar, Y."],"urls":{"Website":"https://arxiv.org/abs/2003.12120,https://ieeexplore.ieee.org/document/9196713/"},"biburl":"https://bibbase.org/service/mendeley/2331788d-b144-3e67-ab8c-4abd7ab569c5","bibbaseid":"soucie-sosik-girdhar-gaussiandirichletrandomfieldsforinferenceoverhighdimensionalcategoricalobservations-2020","role":"author","metadata":{"authorlinks":{"girdhar, y":"https://bibbase.org/service/mendeley/2331788d-b144-3e67-ab8c-4abd7ab569c5"}},"downloads":2},"bibtype":"inproceedings","creationDate":"2020-02-27T19:28:38.342Z","downloads":2,"keywords":[],"search_terms":["gaussian","dirichlet","random","fields","inference","over","high","dimensional","categorical","observations","soucie","sosik","girdhar"],"title":"Gaussian-Dirichlet Random Fields for Inference over High Dimensional Categorical Observations","year":2020,"biburl":"https://bibbase.org/service/mendeley/2331788d-b144-3e67-ab8c-4abd7ab569c5","dataSources":["KqS8yj4KdK5TgR3vo","ya2CyA73rpZseyrZ8","Aba7SDZ3i2zffh3rY","XFrKPG99s5t3W7xuW","2252seNhipfTmjEBQ"]}