Information-theoretic Analysis of Bayesian Test Data Sensitivity

Futoshi Futami, Tomoharu Iwata
Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, PMLR 238:1099-1107, 2024.

Abstract

Bayesian inference is often used to quantify uncertainty. Several recent analyses have rigorously decomposed uncertainty in prediction by Bayesian inference into two types: the inherent randomness in the data generation process and the variability due to lack of data respectively. Existing studies have analyzed these uncertainties from an information-theoretic perspective, assuming the model is well-specified and treating the model parameters as latent variables. However, such information-theoretic uncertainty analysis fails to account for a widely believed property of uncertainty known as sensitivity between test and training data. This means that if the test data is similar to the training data in some sense, the uncertainty will be smaller. In this study, we study such sensitivity using a new decomposition of uncertainty. Our analysis successfully defines such sensitivity using information-theoretic quantities. Furthermore, we extend the existing analysis of Bayesian meta-learning and show the novel sensitivities among tasks for the first time.

Cite this Paper

BibTeX
@InProceedings{pmlr-v238-futami24a, title = { Information-theoretic Analysis of {B}ayesian Test Data Sensitivity }, author = {Futami, Futoshi and Iwata, Tomoharu}, booktitle = {Proceedings of The 27th International Conference on Artificial Intelligence and Statistics}, pages = {1099--1107}, year = {2024}, editor = {Dasgupta, Sanjoy and Mandt, Stephan and Li, Yingzhen}, volume = {238}, series = {Proceedings of Machine Learning Research}, month = {02--04 May}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v238/futami24a/futami24a.pdf}, url = {https://proceedings.mlr.press/v238/futami24a.html}, abstract = { Bayesian inference is often used to quantify uncertainty. Several recent analyses have rigorously decomposed uncertainty in prediction by Bayesian inference into two types: the inherent randomness in the data generation process and the variability due to lack of data respectively. Existing studies have analyzed these uncertainties from an information-theoretic perspective, assuming the model is well-specified and treating the model parameters as latent variables. However, such information-theoretic uncertainty analysis fails to account for a widely believed property of uncertainty known as sensitivity between test and training data. This means that if the test data is similar to the training data in some sense, the uncertainty will be smaller. In this study, we study such sensitivity using a new decomposition of uncertainty. Our analysis successfully defines such sensitivity using information-theoretic quantities. Furthermore, we extend the existing analysis of Bayesian meta-learning and show the novel sensitivities among tasks for the first time. } }
Endnote
%0 Conference Paper %T Information-theoretic Analysis of Bayesian Test Data Sensitivity %A Futoshi Futami %A Tomoharu Iwata %B Proceedings of The 27th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2024 %E Sanjoy Dasgupta %E Stephan Mandt %E Yingzhen Li %F pmlr-v238-futami24a %I PMLR %P 1099--1107 %U https://proceedings.mlr.press/v238/futami24a.html %V 238 %X Bayesian inference is often used to quantify uncertainty. Several recent analyses have rigorously decomposed uncertainty in prediction by Bayesian inference into two types: the inherent randomness in the data generation process and the variability due to lack of data respectively. Existing studies have analyzed these uncertainties from an information-theoretic perspective, assuming the model is well-specified and treating the model parameters as latent variables. However, such information-theoretic uncertainty analysis fails to account for a widely believed property of uncertainty known as sensitivity between test and training data. This means that if the test data is similar to the training data in some sense, the uncertainty will be smaller. In this study, we study such sensitivity using a new decomposition of uncertainty. Our analysis successfully defines such sensitivity using information-theoretic quantities. Furthermore, we extend the existing analysis of Bayesian meta-learning and show the novel sensitivities among tasks for the first time.
APA
Futami, F. & Iwata, T.. (2024). Information-theoretic Analysis of Bayesian Test Data Sensitivity . Proceedings of The 27th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 238:1099-1107 Available from https://proceedings.mlr.press/v238/futami24a.html.

Related Material