MARGIN: Uncovering Deep Neural Networks using Graph Signal Analysis

TL;DR

MARGIN is a versatile graph signal analysis method that improves interpretability of neural networks by identifying influential nodes across various tasks, outperforming existing techniques.

Contribution

The paper introduces MARGIN, a novel graph signal analysis approach that unifies and enhances interpretability methods for neural networks.

Findings

MARGIN outperforms existing interpretability methods across multiple tasks.

It effectively identifies influential nodes that explain model predictions.

The approach is general and adaptable to various interpretability challenges.

Abstract

Interpretability has emerged as a crucial aspect of building trust in machine learning systems, aimed at providing insights into the working of complex neural networks that are otherwise opaque to a user. There are a plethora of existing solutions addressing various aspects of interpretability ranging from identifying prototypical samples in a dataset to explaining image predictions or explaining mis-classifications. While all of these diverse techniques address seemingly different aspects of interpretability, we hypothesize that a large family of interepretability tasks are variants of the same central problem which is identifying \emph{relative} change in a model's prediction. This paper introduces MARGIN, a simple yet general approach to address a large set of interpretability tasks MARGIN exploits ideas rooted in graph signal analysis to determine influential nodes in a graph, which are defined as those nodes that maximally describe a function defined on the graph. By carefully defining task-specific graphs and functions, we demonstrate that MARGIN outperforms existing approaches in a number of disparate interpretability challenges.