The Conditional Entropy Bottleneck
Author:
Open Access
Abstract
Much of the field of Machine Learning exhibits a prominent set of failure modes, including vulnerability to adversarial examples, poor out-of-distribution (OoD) detection, miscalibration, and willingness to memorize random labelings of datasets. We characterize these as failures of robust generalization, which extends the traditional measure of generalization as accuracy or related metrics on a held-out set. We hypothesize that these failures to robustly generalize are due to the learning systems retaining too much information about the training data. To test this hypothesis, we propose the Minimum Necessary Information (MNI) criterion for evaluating the quality of a model. In order to train models that perform well with respect to the MNI criterion, we present a new objective function, the Conditional Entropy Bottleneck (CEB), which is closely related to the Information Bottleneck (IB). We experimentally test our hypothesis by comparing the performance of CEB models with deterministic models and Variational Information Bottleneck (VIB) models on a variety of different datasets and robustness challenges. We find strong empirical evidence supporting our hypothesis that MNI models improve on these problems of robust generalization.Keywords
All Related Versions
This publication has 14 references indexed in Scilit:
- Learnability for the Information BottleneckEntropy, 2019
- Learning Representations for Neural Network-Based Classification Using the Information Bottleneck PrincipleIEEE Transactions on Pattern Analysis and Machine Intelligence, 2019
- Information Dropout: Learning Optimal Representations Through Noisy ComputationIEEE Transactions on Pattern Analysis and Machine Intelligence, 2018
- Towards Evaluating the Robustness of Neural Networks2017
- Strong Data-Processing Inequalities for Channels and Bayesian Networks2017
- Learning and generalization with the information bottleneckTheoretical Computer Science, 2010
- Predictability, Complexity, and LearningNeural Computation, 2001
- Simple statistical gradient-following algorithms for connectionist reinforcement learningMachine Learning, 1992
- A new outlook on Shannon's information measuresIEEE Transactions on Information Theory, 1991
- A Mathematical Theory of CommunicationBell System Technical Journal, 1948