# Information-Theoretic Visual Explanation for Black-Box Classifiers

Source: [Information-Theoretic Visual Explanation for Black-Box Classifiers](<https://www.sota2.com/research/paper/68fa6a611e209131cea27dc9>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-28T13:56:48.909Z

Authors: Jihun Yi, Eunji Kim, Siwon Kim, Sungroh Yoon

Published: 2020-09-23T13:51:16.000Z

## Abstract

In this work, we attempt to explain the prediction of any black-box classifier from an information-theoretic perspective. For each input feature, we compare the classifier outputs with and without that feature using two information-theoretic metrics. Accordingly, we obtain two attribution maps--an information gain (IG) map and a point-wise mutual information (PMI) map. IG map provides a class-independent answer to "How informative is each pixel?", and PMI map offers a class-specific explanation of "How much does each pixel support a specific class?" Compared to existing methods, our method improves the correctness of the attribution maps in terms of a quantitative metric. We also provide a detailed analysis of an ImageNet classifier using the proposed method, and the code is available online.

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2009.11150>)

[PDF](<https://arxiv.org/pdf/2009.11150.pdf>)