Bayesian Image Classification with Deep Convolutional Gaussian Processes

Vincent Dutordoir, Mark Wilk, Artem Artemev, James Hensman
Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, PMLR 108:1529-1539, 2020.

Abstract

In decision-making systems, it is important to have classifiers that have calibrated uncertainties, with an optimisation objective that can be used for automated model selection and training. Gaussian processes (GPs) provide uncertainty estimates and a marginal likelihood objective, but their weak inductive biases lead to inferior accuracy. This has limited their applicability in certain tasks (e.g. image classification). We propose a translation insensitive convolutional kernel, which relaxes the translation invariance constraint imposed by previous convolutional GPs. We show how we can use the marginal likelihood to learn the degree of insensitivity. We also reformulate GP image-to-image convolutional mappings as multi-output GPs, leading to deep convolutional GPs. We show experimentally that our new kernel improves performance in both single-layer and deep models. We also demonstrate that our fully Bayesian approach improves on dropout-based Bayesian deep learning methods in terms of uncertainty and marginal likelihood estimates.

Cite this Paper


BibTeX
@InProceedings{pmlr-v108-dutordoir20a, title = {Bayesian Image Classification with Deep Convolutional Gaussian Processes}, author = {Dutordoir, Vincent and van der Wilk, Mark and Artemev, Artem and Hensman, James}, booktitle = {Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics}, pages = {1529--1539}, year = {2020}, editor = {Silvia Chiappa and Roberto Calandra}, volume = {108}, series = {Proceedings of Machine Learning Research}, month = {26--28 Aug}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v108/dutordoir20a/dutordoir20a.pdf}, url = { http://proceedings.mlr.press/v108/dutordoir20a.html }, abstract = {In decision-making systems, it is important to have classifiers that have calibrated uncertainties, with an optimisation objective that can be used for automated model selection and training. Gaussian processes (GPs) provide uncertainty estimates and a marginal likelihood objective, but their weak inductive biases lead to inferior accuracy. This has limited their applicability in certain tasks (e.g. image classification). We propose a translation insensitive convolutional kernel, which relaxes the translation invariance constraint imposed by previous convolutional GPs. We show how we can use the marginal likelihood to learn the degree of insensitivity. We also reformulate GP image-to-image convolutional mappings as multi-output GPs, leading to deep convolutional GPs. We show experimentally that our new kernel improves performance in both single-layer and deep models. We also demonstrate that our fully Bayesian approach improves on dropout-based Bayesian deep learning methods in terms of uncertainty and marginal likelihood estimates.} }
Endnote
%0 Conference Paper %T Bayesian Image Classification with Deep Convolutional Gaussian Processes %A Vincent Dutordoir %A Mark Wilk %A Artem Artemev %A James Hensman %B Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2020 %E Silvia Chiappa %E Roberto Calandra %F pmlr-v108-dutordoir20a %I PMLR %P 1529--1539 %U http://proceedings.mlr.press/v108/dutordoir20a.html %V 108 %X In decision-making systems, it is important to have classifiers that have calibrated uncertainties, with an optimisation objective that can be used for automated model selection and training. Gaussian processes (GPs) provide uncertainty estimates and a marginal likelihood objective, but their weak inductive biases lead to inferior accuracy. This has limited their applicability in certain tasks (e.g. image classification). We propose a translation insensitive convolutional kernel, which relaxes the translation invariance constraint imposed by previous convolutional GPs. We show how we can use the marginal likelihood to learn the degree of insensitivity. We also reformulate GP image-to-image convolutional mappings as multi-output GPs, leading to deep convolutional GPs. We show experimentally that our new kernel improves performance in both single-layer and deep models. We also demonstrate that our fully Bayesian approach improves on dropout-based Bayesian deep learning methods in terms of uncertainty and marginal likelihood estimates.
APA
Dutordoir, V., Wilk, M., Artemev, A. & Hensman, J.. (2020). Bayesian Image Classification with Deep Convolutional Gaussian Processes. Proceedings of the Twenty Third International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 108:1529-1539 Available from http://proceedings.mlr.press/v108/dutordoir20a.html .

Related Material