The Confusion is Real: GRAPHIC -- A Network Science Approach to Confusion Matrices in Deep Learning

TL;DR

GRAPHIC is a network science-based method that visualizes and analyzes class confusion in neural networks across training, providing insights into learning dynamics and dataset issues.

Contribution

It introduces a systematic, architecture-agnostic approach using confusion matrices as graphs to understand class confusion and learning behavior.

Findings

Reveals class similarities like flatfish and man.

Identifies labeling ambiguities validated by human study.

Provides insights into dataset issues and neural network behavior.

Abstract

Explainable artificial intelligence has emerged as a promising field of research to address reliability concerns in artificial intelligence. Despite significant progress in explainable artificial intelligence, few methods provide a systematic way to visualize and understand how classes are confused and how their relationships evolve as training progresses. In this work, we present GRAPHIC, an architecture-agnostic approach that analyzes neural networks on a class level. It leverages confusion matrices derived from intermediate layers using linear classifiers. We interpret these as adjacency matrices of directed graphs, allowing tools from network science to visualize and quantify learning dynamics across training epochs and intermediate layers. GRAPHIC provides insights into linear class separability, dataset issues, and architectural behavior, revealing, for example, similarities between flatfish and man and labeling ambiguities validated in a human study. In summary, by uncovering real confusions, GRAPHIC offers new perspectives on how neural networks learn. The code is available at https://github.com/Johanna-S-Froehlich/GRAPHIC.