Variational zero-inflated Gaussian processes with sparse kernels

Pashupati Hegde, Markus Heinonen, Samuel Kaski
Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence, PMLR R16:360-370, 2018.

Abstract

Zero-inflated datasets, which have an excess of zero outputs, are commonly encountered in problems such as climate or rare event modelling. Conventional machine learning approaches tend to overestimate the non-zeros leading to poor performance. We propose a novel model family of zero-inflated Gaussian processes (ZiGP) for such zero-inflated datasets, produced by sparse kernels through learning a latent probit Gaussian process that can zero out kernel rows and columns whenever the signal is absent. The ZiGPs are particularly useful for making the powerful Gaussian process networks more interpretable. We introduce sparse GP networks where variable-order latent modelling is achieved through sparse mixing signals. We derive the non-trivial stochastic variational inference tractably for scalable learning of the sparse kernels in both models. The novel output-sparse approach improves both prediction of zero-inflated data and interpretability of latent mixing models.

Cite this Paper

BibTeX
@InProceedings{pmlr-vR16-hegde18a, title = {Variational zero-inflated {G}aussian processes with sparse kernels}, author = {Hegde, Pashupati and Heinonen, Markus and Kaski, Samuel}, booktitle = {Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence}, pages = {360--370}, year = {2018}, editor = {Globerson, Amir and Silva, Ricardo}, volume = {R16}, series = {Proceedings of Machine Learning Research}, month = {06--10 Aug}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/r16/main/assets/hegde18a/hegde18a.pdf}, url = {https://proceedings.mlr.press/r16/hegde18a.html}, abstract = {Zero-inflated datasets, which have an excess of zero outputs, are commonly encountered in problems such as climate or rare event modelling. Conventional machine learning approaches tend to overestimate the non-zeros leading to poor performance. We propose a novel model family of zero-inflated Gaussian processes (ZiGP) for such zero-inflated datasets, produced by sparse kernels through learning a latent probit Gaussian process that can zero out kernel rows and columns whenever the signal is absent. The ZiGPs are particularly useful for making the powerful Gaussian process networks more interpretable. We introduce sparse GP networks where variable-order latent modelling is achieved through sparse mixing signals. We derive the non-trivial stochastic variational inference tractably for scalable learning of the sparse kernels in both models. The novel output-sparse approach improves both prediction of zero-inflated data and interpretability of latent mixing models.}, note = {Reissued by PMLR on 04 October 2026.} }
Endnote
%0 Conference Paper %T Variational zero-inflated Gaussian processes with sparse kernels %A Pashupati Hegde %A Markus Heinonen %A Samuel Kaski %B Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2018 %E Amir Globerson %E Ricardo Silva %F pmlr-vR16-hegde18a %I PMLR %P 360--370 %U https://proceedings.mlr.press/r16/hegde18a.html %V R16 %X Zero-inflated datasets, which have an excess of zero outputs, are commonly encountered in problems such as climate or rare event modelling. Conventional machine learning approaches tend to overestimate the non-zeros leading to poor performance. We propose a novel model family of zero-inflated Gaussian processes (ZiGP) for such zero-inflated datasets, produced by sparse kernels through learning a latent probit Gaussian process that can zero out kernel rows and columns whenever the signal is absent. The ZiGPs are particularly useful for making the powerful Gaussian process networks more interpretable. We introduce sparse GP networks where variable-order latent modelling is achieved through sparse mixing signals. We derive the non-trivial stochastic variational inference tractably for scalable learning of the sparse kernels in both models. The novel output-sparse approach improves both prediction of zero-inflated data and interpretability of latent mixing models. %Z Reissued by PMLR on 04 October 2026.
APA
Hegde, P., Heinonen, M. & Kaski, S.. (2018). Variational zero-inflated Gaussian processes with sparse kernels. Proceedings of the 34th Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research R16:360-370 Available from https://proceedings.mlr.press/r16/hegde18a.html. Reissued by PMLR on 04 October 2026.

Related Material