Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces

William Herlands, Andrew Wilson, Hannes Nickisch, Seth Flaxman, Daniel Neill, Wilbert Van Panhuis, Eric Xing
Proceedings of the 19th International Conference on Artificial Intelligence and Statistics, PMLR 51:1013-1021, 2016.

Abstract

We present a scalable Gaussian process model for identifying and characterizing smooth multidimensional changepoints, and automatically learning changes in expressive covariance structure. We use Random Kitchen Sink features to flexibly define a change surface in combination with expressive spectral mixture kernels to capture the complex statistical structure. Finally, through the use of novel methods for additive non-separable kernels, we can scale the model to large datasets. We demonstrate the model on numerical and real world data, including a large spatio-temporal disease dataset where we identify previously unknown heterogeneous changes in space and time.

Cite this Paper

BibTeX
@InProceedings{pmlr-v51-herlands16, title = {Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces}, author = {Herlands, William and Wilson, Andrew and Nickisch, Hannes and Flaxman, Seth and Neill, Daniel and Van Panhuis, Wilbert and Xing, Eric}, booktitle = {Proceedings of the 19th International Conference on Artificial Intelligence and Statistics}, pages = {1013--1021}, year = {2016}, editor = {Gretton, Arthur and Robert, Christian C.}, volume = {51}, series = {Proceedings of Machine Learning Research}, address = {Cadiz, Spain}, month = {09--11 May}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v51/herlands16.pdf}, url = {https://proceedings.mlr.press/v51/herlands16.html}, abstract = {We present a scalable Gaussian process model for identifying and characterizing smooth multidimensional changepoints, and automatically learning changes in expressive covariance structure. We use Random Kitchen Sink features to flexibly define a change surface in combination with expressive spectral mixture kernels to capture the complex statistical structure. Finally, through the use of novel methods for additive non-separable kernels, we can scale the model to large datasets. We demonstrate the model on numerical and real world data, including a large spatio-temporal disease dataset where we identify previously unknown heterogeneous changes in space and time.} }
Endnote
%0 Conference Paper %T Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces %A William Herlands %A Andrew Wilson %A Hannes Nickisch %A Seth Flaxman %A Daniel Neill %A Wilbert Van Panhuis %A Eric Xing %B Proceedings of the 19th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2016 %E Arthur Gretton %E Christian C. Robert %F pmlr-v51-herlands16 %I PMLR %P 1013--1021 %U https://proceedings.mlr.press/v51/herlands16.html %V 51 %X We present a scalable Gaussian process model for identifying and characterizing smooth multidimensional changepoints, and automatically learning changes in expressive covariance structure. We use Random Kitchen Sink features to flexibly define a change surface in combination with expressive spectral mixture kernels to capture the complex statistical structure. Finally, through the use of novel methods for additive non-separable kernels, we can scale the model to large datasets. We demonstrate the model on numerical and real world data, including a large spatio-temporal disease dataset where we identify previously unknown heterogeneous changes in space and time.
RIS
TY - CPAPER TI - Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces AU - William Herlands AU - Andrew Wilson AU - Hannes Nickisch AU - Seth Flaxman AU - Daniel Neill AU - Wilbert Van Panhuis AU - Eric Xing BT - Proceedings of the 19th International Conference on Artificial Intelligence and Statistics DA - 2016/05/02 ED - Arthur Gretton ED - Christian C. Robert ID - pmlr-v51-herlands16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 51 SP - 1013 EP - 1021 L1 - http://proceedings.mlr.press/v51/herlands16.pdf UR - https://proceedings.mlr.press/v51/herlands16.html AB - We present a scalable Gaussian process model for identifying and characterizing smooth multidimensional changepoints, and automatically learning changes in expressive covariance structure. We use Random Kitchen Sink features to flexibly define a change surface in combination with expressive spectral mixture kernels to capture the complex statistical structure. Finally, through the use of novel methods for additive non-separable kernels, we can scale the model to large datasets. We demonstrate the model on numerical and real world data, including a large spatio-temporal disease dataset where we identify previously unknown heterogeneous changes in space and time. ER -
APA
Herlands, W., Wilson, A., Nickisch, H., Flaxman, S., Neill, D., Van Panhuis, W. & Xing, E.. (2016). Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces. Proceedings of the 19th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 51:1013-1021 Available from https://proceedings.mlr.press/v51/herlands16.html.

Related Material