Calibrating Deep Convolutional Gaussian Processes

Gia-Lac Tran, Edwin V. Bonilla, John Cunningham, Pietro Michiardi, Maurizio Filippone
Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics, PMLR 89:1554-1563, 2019.

Abstract

The wide adoption of Convolutional Neural Networks CNNs in applications where decision-making under uncertainty is fundamental, has brought a great deal of attention to the ability of these models to accurately quantify the uncertainty in their predictions. Previous work on combining CNNs with Gaussian processes GPs has been developed under the assumption that the predictive probabilities of these models are well-calibrated. In this paper we show that, in fact, current combinations of CNNs and GPs are miscalibrated. We proposes a novel combination that considerably outperforms previous approaches on this aspect, while achieving state-of-the-art performance on image classification tasks.

Cite this Paper

BibTeX
@InProceedings{pmlr-v89-tran19a, title = {Calibrating Deep Convolutional Gaussian Processes}, author = {Tran, Gia-Lac and Bonilla, Edwin V. and Cunningham, John and Michiardi, Pietro and Filippone, Maurizio}, booktitle = {Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics}, pages = {1554--1563}, year = {2019}, editor = {Chaudhuri, Kamalika and Sugiyama, Masashi}, volume = {89}, series = {Proceedings of Machine Learning Research}, month = {16--18 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v89/tran19a/tran19a.pdf}, url = {https://proceedings.mlr.press/v89/tran19a.html}, abstract = {The wide adoption of Convolutional Neural Networks CNNs in applications where decision-making under uncertainty is fundamental, has brought a great deal of attention to the ability of these models to accurately quantify the uncertainty in their predictions. Previous work on combining CNNs with Gaussian processes GPs has been developed under the assumption that the predictive probabilities of these models are well-calibrated. In this paper we show that, in fact, current combinations of CNNs and GPs are miscalibrated. We proposes a novel combination that considerably outperforms previous approaches on this aspect, while achieving state-of-the-art performance on image classification tasks.} }
Endnote
%0 Conference Paper %T Calibrating Deep Convolutional Gaussian Processes %A Gia-Lac Tran %A Edwin V. Bonilla %A John Cunningham %A Pietro Michiardi %A Maurizio Filippone %B Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Masashi Sugiyama %F pmlr-v89-tran19a %I PMLR %P 1554--1563 %U https://proceedings.mlr.press/v89/tran19a.html %V 89 %X The wide adoption of Convolutional Neural Networks CNNs in applications where decision-making under uncertainty is fundamental, has brought a great deal of attention to the ability of these models to accurately quantify the uncertainty in their predictions. Previous work on combining CNNs with Gaussian processes GPs has been developed under the assumption that the predictive probabilities of these models are well-calibrated. In this paper we show that, in fact, current combinations of CNNs and GPs are miscalibrated. We proposes a novel combination that considerably outperforms previous approaches on this aspect, while achieving state-of-the-art performance on image classification tasks.
APA
Tran, G., Bonilla, E.V., Cunningham, J., Michiardi, P. & Filippone, M.. (2019). Calibrating Deep Convolutional Gaussian Processes. Proceedings of the Twenty-Second International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 89:1554-1563 Available from https://proceedings.mlr.press/v89/tran19a.html.

Related Material