Axiomatic Attribution for Deep Networks

Mukund Sundararajan, Ankur Taly, Qiqi Yan
3/4/2017
cs.LG

Official code: ankurtaly/Integrated-Gradients (653 stars, last commit 2022-02-23).

Abstract

We study the problem of attributing the prediction of a deep network to its input features, a problem previously studied by several other works. We identify two fundamental axioms---Sensitivity and Implementation Invariance that attribution methods ought to satisfy. We show that they are not satisfied by most known attribution methods, which we consider to be a fundamental weakness of those methods. We use the axioms to guide the design of a new attribution method called Integrated Gradients. Our method requires no modification to the original network and is extremely simple to implement; it just needs a few calls to the standard gradient operator. We apply this method to a couple of image models, a couple of text models and a chemistry model, demonstrating its ability to debug networks, to extract rules from a network, and to enable users to engage with models better.

View on arXivView PDF

Code Implementations(1)

ankurtaly/Integrated-GradientsOfficial100%

Attributing predictions made by the Inception network using the Integrated Gradients method

653108Python, Jupyter NotebookNov 4, 20164 years ago

Similar Papers

Feature Attribution from First Principles

87% match1 repo42025

Improving performance of deep learning models with axiomatic attribution priors and expected gradients

87% match1 repo1282019

Towards better understanding of gradient-based attribution methods for Deep Neural Networks

87% match1 repo8822017

A Polynomial Architecture-Attribution Co-Design Framework for Exact Aumann-Shapley Attribution in GNNs

86% match2026

Beyond Single Path Integrated Gradients for Reliable Input Attribution via Randomized Path Sampling

86% match2023

Shedding Light on Depth: Explainability Assessment in Monocular Depth Estimation

85% match2025

Cite this paper

@article{sundararajan2017axiomatic,
  title  = {Axiomatic Attribution for Deep Networks},
  author = {Mukund Sundararajan and Ankur Taly and Qiqi Yan},
  year   = {2017},
  eprint = {1703.01365},
  archivePrefix = {arXiv},
  url    = {https://arxiv.org/abs/1703.01365}
}

Add a badge to your README

Shows whether official code is linked for this paper.

Code of Paper badge
[![Code of Paper](https://codeofpaper.com/badge/1703.01365.svg)](https://codeofpaper.com/paper/1703.01365)

Discussion