Uncertainty-aware sensitivity analysis using Rényi divergences

Topi Paananen, Michael Riis Andersen, Aki Vehtari
Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence, PMLR 161:1185-1194, 2021.

Abstract

For nonlinear supervised learning models, assessing the importance of predictor variables or their interactions is not straightforward because importance can vary in the domain of the variables. Importance can be assessed locally with sensitivity analysis using general methods that rely on the model’s predictions or their derivatives. In this work, we extend derivative based sensitivity analysis to a Bayesian setting by differentiating the R\’{e}nyi divergence of a model’s predictive distribution. By utilising the predictive distribution instead of a point prediction, the model uncertainty is taken into account in a principled way. Our empirical results on simulated and real data sets demonstrate accurate and reliable identification of important variables and interaction effects compared to alternative methods.

Cite this Paper

BibTeX
@InProceedings{pmlr-v161-paananen21a, title = {Uncertainty-aware sensitivity analysis using Rényi divergences}, author = {Paananen, Topi and Andersen, Michael Riis and Vehtari, Aki}, booktitle = {Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence}, pages = {1185--1194}, year = {2021}, editor = {de Campos, Cassio and Maathuis, Marloes H.}, volume = {161}, series = {Proceedings of Machine Learning Research}, month = {27--30 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v161/paananen21a/paananen21a.pdf}, url = {https://proceedings.mlr.press/v161/paananen21a.html}, abstract = {For nonlinear supervised learning models, assessing the importance of predictor variables or their interactions is not straightforward because importance can vary in the domain of the variables. Importance can be assessed locally with sensitivity analysis using general methods that rely on the model’s predictions or their derivatives. In this work, we extend derivative based sensitivity analysis to a Bayesian setting by differentiating the R\’{e}nyi divergence of a model’s predictive distribution. By utilising the predictive distribution instead of a point prediction, the model uncertainty is taken into account in a principled way. Our empirical results on simulated and real data sets demonstrate accurate and reliable identification of important variables and interaction effects compared to alternative methods.} }
Endnote
%0 Conference Paper %T Uncertainty-aware sensitivity analysis using Rényi divergences %A Topi Paananen %A Michael Riis Andersen %A Aki Vehtari %B Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence %C Proceedings of Machine Learning Research %D 2021 %E Cassio de Campos %E Marloes H. Maathuis %F pmlr-v161-paananen21a %I PMLR %P 1185--1194 %U https://proceedings.mlr.press/v161/paananen21a.html %V 161 %X For nonlinear supervised learning models, assessing the importance of predictor variables or their interactions is not straightforward because importance can vary in the domain of the variables. Importance can be assessed locally with sensitivity analysis using general methods that rely on the model’s predictions or their derivatives. In this work, we extend derivative based sensitivity analysis to a Bayesian setting by differentiating the R\’{e}nyi divergence of a model’s predictive distribution. By utilising the predictive distribution instead of a point prediction, the model uncertainty is taken into account in a principled way. Our empirical results on simulated and real data sets demonstrate accurate and reliable identification of important variables and interaction effects compared to alternative methods.
APA
Paananen, T., Andersen, M.R. & Vehtari, A.. (2021). Uncertainty-aware sensitivity analysis using Rényi divergences. Proceedings of the Thirty-Seventh Conference on Uncertainty in Artificial Intelligence, in Proceedings of Machine Learning Research 161:1185-1194 Available from https://proceedings.mlr.press/v161/paananen21a.html.

Related Material