Explaining, Fast and Slow: Abstraction and Refinement of Provable Explanations

TL;DR

This paper introduces an abstraction-refinement method that efficiently computes provably sufficient explanations for neural network predictions, balancing formal guarantees with scalability and interpretability.

Contribution

It proposes a novel abstraction-refinement technique that reduces neural network size for faster provable explanations, with iterative refinement for accuracy.

Findings

Significantly improves explanation computation speed

Provides fine-grained interpretation across abstraction levels

Maintains formal guarantees of explanation sufficiency

Abstract

Despite significant advancements in post-hoc explainability techniques for neural networks, many current methods rely on heuristics and do not provide formally provable guarantees over the explanations provided. Recent work has shown that it is possible to obtain explanations with formal guarantees by identifying subsets of input features that are sufficient to determine that predictions remain unchanged using neural network verification techniques. Despite the appeal of these explanations, their computation faces significant scalability challenges. In this work, we address this gap by proposing a novel abstraction-refinement technique for efficiently computing provably sufficient explanations of neural network predictions. Our method abstracts the original large neural network by constructing a substantially reduced network, where a sufficient explanation of the reduced network is also provably sufficient for the original network, hence significantly speeding up the verification process. If the explanation is in sufficient on the reduced network, we iteratively refine the network size by gradually increasing it until convergence. Our experiments demonstrate that our approach enhances the efficiency of obtaining provably sufficient explanations for neural network predictions while additionally providing a fine-grained interpretation of the network's predictions across different abstraction levels.