Post-hoc Part-prototype Networks

TL;DR

This paper introduces a novel post-hoc part-prototype network that decomposes a trained model's classification head into interpretable prototypes, providing accurate, faithful explanations of both where and what the model focuses on, without sacrificing performance.

Contribution

It presents the first post-hoc part-prototype network that guarantees high accuracy while offering interpretable explanations by decomposing the classification head into part-prototypes.

Findings

Provides more faithful explanations than prior methods.

Achieves better part-prototypes quantitatively.

Maintains model performance with interpretable prototypes.

Abstract

Post-hoc explainability methods such as Grad-CAM are popular because they do not influence the performance of a trained model. However, they mainly reveal "where" a model looks at for a given input, fail to explain "what" the model looks for (e.g., what is important to classify a bird image to a Scott Oriole?). Existing part-prototype networks leverage part-prototypes (e.g., characteristic Scott Oriole's wing and head) to answer both "where" and "what", but often under-perform their black box counterparts in the accuracy. Therefore, a natural question is: can one construct a network that answers both "where" and "what" in a post-hoc manner to guarantee the model's performance? To this end, we propose the first post-hoc part-prototype network via decomposing the classification head of a trained model into a set of interpretable part-prototypes. Concretely, we propose an unsupervised prototype discovery and refining strategy to obtain prototypes that can precisely reconstruct the classification head, yet being interpretable. Besides guaranteeing the performance, we show that our network offers more faithful explanations qualitatively and yields even better part-prototypes quantitatively than prior part-prototype networks.