Local-to-Global Logical Explanations for Deep Vision Models

TL;DR

This paper introduces logical explanation methods for deep vision models that generate human-understandable explanations in the form of primitive concepts, providing both local and global interpretability with high fidelity.

Contribution

It presents a novel approach to generate logical, human-recognizable explanations for deep models in vision, covering local, global, and multi-class explanations.

Findings

Explanations are highly faithful and cover the model's behavior.

Logical explanations are effective across challenging vision datasets.

The method supports explanations for multiple classes simultaneously.

Abstract

While deep neural networks are extremely effective at classifying images, they remain opaque and hard to interpret. We introduce local and global explanation methods for black-box models that generate explanations in terms of human-recognizable primitive concepts. Both the local explanations for a single image and the global explanations for a set of images are cast as logical formulas in monotone disjunctive-normal-form (MDNF), whose satisfaction guarantees that the model yields a high score on a given class. We also present an algorithm for explaining the classification of examples into multiple classes in the form of a monotone explanation list over primitive concepts. Despite their simplicity and interpretability we show that the explanations maintain high fidelity and coverage with respect to the blackbox models they seek to explain in challenging vision datasets.