Understanding Classifiers with Generative Models

Laëtitia Shao, Yang Song, Stefano Ermon

Keywords: OOD detection, adversarial samples detection, deep learning, classification

Abstract:

Although deep neural networks are effective on supervised learning tasks, they have been shown to be brittle. They are prone to overfitting on their training distribution and are easily fooled by small adversarial perturbations. In this paper, we leverage generative models to identify and characterize instances where classifiers fail to generalize.
We propose a generative model of the features extracted by a classifier, and show using rigorous hypothesis testing that errors tend to occur when features are assigned low-probability by our model. From this observation, we develop a detection criteria that we test against different sources of classification mistakes: mistakes made on the test set due to poor model generalization, adversarial samples and out-of-distribution samples. Our approach is agnostic to class labels from the training set which makes it applicable to models trained in a semi-supervised way.

One-sentence Summary: We show a new characterization of classification mistakes using a generative model learned on the feature space

Code Of Ethics: I acknowledge that I and all co-authors of this work have read and commit to adhering to the ICLR Code of Ethics

Reviewed Version (pdf): https://openreview.net/references/pdf?id=oLWOrRuL19