Improving Prototypical Visual Explanations with Reward Reweighing, Reselection, and Retraining

TL;DR

This paper introduces the R3 framework that enhances the interpretability and accuracy of ProtoPNet by using human feedback to reweigh, reseleect, and retrain the model post hoc.

Contribution

The paper presents a novel offline post-processing method that improves prototype alignment with human preferences and boosts interpretability and accuracy of ProtoPNet.

Findings

R3 improves interpretability of ProtoPNet.

R3 enhances predictive accuracy.

R3 aligns prototypes with human preferences.

Abstract

In recent years, work has gone into developing deep interpretable methods for image classification that clearly attributes a model's output to specific features of the data. One such of these methods is the Prototypical Part Network (ProtoPNet), which attempts to classify images based on meaningful parts of the input. While this architecture is able to produce visually interpretable classifications, it often learns to classify based on parts of the image that are not semantically meaningful. To address this problem, we propose the Reward Reweighing, Reselecting, and Retraining (R3) post-processing framework, which performs three additional corrective updates to a pretrained ProtoPNet in an offline and efficient manner. The first two steps involve learning a reward model based on collected human feedback and then aligning the prototypes with human preferences. The final step is retraining, which realigns the base features and the classifier layer of the original model with the updated prototypes. We find that our R3 framework consistently improves both the interpretability and the predictive accuracy of ProtoPNet and its variants.