Understanding CNNs from excitations

TL;DR

This paper introduces a gradient-free method for explaining CNN decisions by extracting positive and negative excitations layer-by-layer, leading to improved saliency maps and interpretability.

Contribution

It proposes a novel positive and negative excitation concept and a double-chain backpropagation method to enhance CNN interpretability without relying on gradients.

Findings

Outperforms state-of-the-art saliency methods in pixel removal tasks

Effectively guides adversarial perturbation generation

Shows strong correlation between positive and negative excitations

Abstract

Saliency maps have proven to be a highly efficacious approach for explicating the decisions of Convolutional Neural Networks. However, extant methodologies predominantly rely on gradients, which constrain their ability to explicate complex models. Furthermore, such approaches are not fully adept at leveraging negative gradient information to improve interpretive veracity. In this study, we present a novel concept, termed positive and negative excitation, which enables the direct extraction of positive and negative excitation for each layer, thus enabling complete layer-by-layer information utilization sans gradients. To organize these excitations into final saliency maps, we introduce a double-chain backpropagation procedure. A comprehensive experimental evaluation, encompassing both binary classification and multi-classification tasks, was conducted to gauge the effectiveness of our proposed method. Encouragingly, the results evince that our approach offers a significant improvement over the state-of-the-art methods in terms of salient pixel removal, minor pixel removal, and inconspicuous adversarial perturbation generation guidance. Additionally, we verify the correlation between positive and negative excitations.