# Explaining Deep Neural Networks with a Polynomial Time Algorithm for Shapley Values Approximation

## TL;DR

This paper introduces a polynomial-time algorithm for approximating Shapley values to explain deep neural networks, providing a theoretically grounded and efficient alternative to existing attribution methods.

## Contribution

It presents a novel polynomial-time approximation method for Shapley values in deep neural networks, improving explanation accuracy with strong theoretical backing.

## Key findings

- Our method outperforms existing attribution techniques in approximation accuracy.
- The approach is computationally efficient, enabling practical use in large networks.
- It offers a theoretically justified framework for neural network explanations.

## Abstract

The problem of explaining the behavior of deep neural networks has recently gained a lot of attention. While several attribution methods have been proposed, most come without strong theoretical foundations, which raises questions about their reliability. On the other hand, the literature on cooperative game theory suggests Shapley values as a unique way of assigning relevance scores such that certain desirable properties are satisfied. Unfortunately, the exact evaluation of Shapley values is prohibitively expensive, exponential in the number of input features. In this work, by leveraging recent results on uncertainty propagation, we propose a novel, polynomial-time approximation of Shapley values in deep neural networks. We show that our method produces significantly better approximations of Shapley values than existing state-of-the-art attribution methods.