Visualization of Supervised and Self-Supervised Neural Networks via Attribution Guided Factorization

TL;DR

This paper introduces a novel visualization algorithm for neural networks that enhances class-specific explanations by integrating gradient and attribution methods, overcoming saliency bias, and demonstrating semantic learning in self-supervised models.

Contribution

The work develops a new explainability algorithm that provides per-class visualizations, correcting saliency bias and enabling insights into self-supervised learning.

Findings

Achieves state-of-the-art results on gradient-based and attribution benchmarks.

Effectively visualizes class-specific features beyond the predicted label.

Demonstrates that self-supervised models learn semantic information.

Abstract

Neural network visualization techniques mark image locations by their relevancy to the network's classification. Existing methods are effective in highlighting the regions that affect the resulting classification the most. However, as we show, these methods are limited in their ability to identify the support for alternative classifications, an effect we name {\em the saliency bias} hypothesis. In this work, we integrate two lines of research: gradient-based methods and attribution-based methods, and develop an algorithm that provides per-class explainability. The algorithm back-projects the per pixel local influence, in a manner that is guided by the local attributions, while correcting for salient features that would otherwise bias the explanation. In an extensive battery of experiments, we demonstrate the ability of our methods to class-specific visualization, and not just the predicted label. Remarkably, the method obtains state of the art results in benchmarks that are commonly applied to gradient-based methods as well as in those that are employed mostly for evaluating attribution methods. Using a new unsupervised procedure, our method is also successful in demonstrating that self-supervised methods learn semantic information.