Bounded logit attention: Learning to explain image classifiers

TL;DR

This paper introduces Bounded Logit Attention (BLA), a trainable module for convolutional image classifiers that provides scalable, variable-sized explanations, outperforming existing methods in user preference and applicability.

Contribution

BLA is a novel, scalable, and modular explanation method that improves upon prior feature selection techniques for image classification.

Findings

BLA scales to real-world image classification tasks.

BLA explanations are preferred over Grad-CAM in user studies.

BLA can be used as a post-hoc explanation tool or integrated during training.

Abstract

Explainable artificial intelligence is the attempt to elucidate the workings of systems too complex to be directly accessible to human cognition through suitable side-information referred to as "explanations". We present a trainable explanation module for convolutional image classifiers we call bounded logit attention (BLA). The BLA module learns to select a subset of the convolutional feature map for each input instance, which then serves as an explanation for the classifier's prediction. BLA overcomes several limitations of the instancewise feature selection method "learning to explain" (L2X) introduced by Chen et al. (2018): 1) BLA scales to real-world sized image classification problems, and 2) BLA offers a canonical way to learn explanations of variable size. Due to its modularity BLA lends itself to transfer learning setups and can also be employed as a post-hoc add-on to trained classifiers. Beyond explainability, BLA may serve as a general purpose method for differentiable approximation of subset selection. In a user study we find that BLA explanations are preferred over explanations generated by the popular (Grad-)CAM method.