Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces
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
Text
Herlands et al. "Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces." International Conference on Artificial Intelligence and Statistics, 2016.Markdown
[Herlands et al. "Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces." International Conference on Artificial Intelligence and Statistics, 2016.](https://mlanthology.org/aistats/2016/herlands2016aistats-scalable/)BibTeX
@inproceedings{herlands2016aistats-scalable,
title = {{Scalable Gaussian Processes for Characterizing Multidimensional Change Surfaces}},
author = {Herlands, William and Wilson, Andrew Gordon and Nickisch, Hannes and Flaxman, Seth R. and Neill, Daniel B. and Van Panhuis, Wilbert and Xing, Eric P.},
booktitle = {International Conference on Artificial Intelligence and Statistics},
year = {2016},
pages = {1013-1021},
url = {https://mlanthology.org/aistats/2016/herlands2016aistats-scalable/}
}