QPM: Discrete Optimization for Globally Interpretable Image Classification

TL;DR

QPM introduces a novel discrete optimization approach to learn globally interpretable class representations in deep neural networks, enhancing understanding of model behavior while maintaining high accuracy across various datasets.

Contribution

The paper presents QPM, a method that learns compact, contrastive class representations using discrete optimization, enabling global interpretability in image classification models.

Findings

QPM achieves state-of-the-art interpretability on multiple datasets.

QPM maintains high classification accuracy comparable to non-interpretable models.

QPM provides clear, contrastive class representations that are easy to compare.

Abstract

Understanding the classifications of deep neural networks, e.g. used in safety-critical situations, is becoming increasingly important. While recent models can locally explain a single decision, to provide a faithful global explanation about an accurate model's general behavior is a more challenging open task. Towards that goal, we introduce the Quadratic Programming Enhanced Model (QPM), which learns globally interpretable class representations. QPM represents every class with a binary assignment of very few, typically 5, features, that are also assigned to other classes, ensuring easily comparable contrastive class representations. This compact binary assignment is found using discrete optimization based on predefined similarity measures and interpretability constraints. The resulting optimal assignment is used to fine-tune the diverse features, so that each of them becomes the shared general concept between the assigned classes. Extensive evaluations show that QPM delivers unprecedented global interpretability across small and large-scale datasets while setting the state of the art for the accuracy of interpretable models.