Scalable gaussian processes for characterizing multidimensional change surfaces
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 exibly define a change surface in combination with expressive spectral m...
| Main Authors: | , , , , , , |
|---|---|
| Format: | Conference item |
| Published: |
Journal of Machine Learning Research
2016
|
| _version_ | 1797084374868426752 |
|---|---|
| author | Herlands, W Wilson, A Nickisch, H Flaxman, S Neill, D van Panhuis, W Xing, E |
| author_facet | Herlands, W Wilson, A Nickisch, H Flaxman, S Neill, D van Panhuis, W Xing, E |
| author_sort | Herlands, W |
| collection | OXFORD |
| description | 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 exibly 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. |
| first_indexed | 2024-03-07T01:54:32Z |
| format | Conference item |
| id | oxford-uuid:9b3fde87-d9fb-404f-9be3-3a13b2fb4afd |
| institution | University of Oxford |
| last_indexed | 2024-03-07T01:54:32Z |
| publishDate | 2016 |
| publisher | Journal of Machine Learning Research |
| record_format | dspace |
| spelling | oxford-uuid:9b3fde87-d9fb-404f-9be3-3a13b2fb4afd2022-03-27T00:27:30ZScalable gaussian processes for characterizing multidimensional change surfacesConference itemhttp://purl.org/coar/resource_type/c_5794uuid:9b3fde87-d9fb-404f-9be3-3a13b2fb4afdSymplectic Elements at OxfordJournal of Machine Learning Research2016Herlands, WWilson, ANickisch, HFlaxman, SNeill, Dvan Panhuis, WXing, EWe 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 exibly 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. |
| spellingShingle | Herlands, W Wilson, A Nickisch, H Flaxman, S Neill, D van Panhuis, W Xing, E Scalable gaussian processes for characterizing multidimensional change surfaces |
| title | Scalable gaussian processes for characterizing multidimensional change surfaces |
| title_full | Scalable gaussian processes for characterizing multidimensional change surfaces |
| title_fullStr | Scalable gaussian processes for characterizing multidimensional change surfaces |
| title_full_unstemmed | Scalable gaussian processes for characterizing multidimensional change surfaces |
| title_short | Scalable gaussian processes for characterizing multidimensional change surfaces |
| title_sort | scalable gaussian processes for characterizing multidimensional change surfaces |
| work_keys_str_mv | AT herlandsw scalablegaussianprocessesforcharacterizingmultidimensionalchangesurfaces AT wilsona scalablegaussianprocessesforcharacterizingmultidimensionalchangesurfaces AT nickischh scalablegaussianprocessesforcharacterizingmultidimensionalchangesurfaces AT flaxmans scalablegaussianprocessesforcharacterizingmultidimensionalchangesurfaces AT neilld scalablegaussianprocessesforcharacterizingmultidimensionalchangesurfaces AT vanpanhuisw scalablegaussianprocessesforcharacterizingmultidimensionalchangesurfaces AT xinge scalablegaussianprocessesforcharacterizingmultidimensionalchangesurfaces |