Neuron Shapley: Discovering the Responsible Neurons

TL;DR

Neuron Shapley introduces a novel method to quantify individual neuron contributions in deep networks, effectively identifying critical filters that influence predictions, biases, and vulnerabilities, thereby enhancing interpretability and model repair.

Contribution

This work presents Neuron Shapley, a new framework that accounts for neuron interactions to identify important neurons, along with a multi-arm bandit algorithm for efficient Shapley value estimation.

Findings

Removing top filters destroys Inception-v3 accuracy.

Filters responsible for biases and adversarial vulnerabilities identified.

Neuron Shapley outperforms activation-based importance methods.

Abstract

We develop Neuron Shapley as a new framework to quantify the contribution of individual neurons to the prediction and performance of a deep network. By accounting for interactions across neurons, Neuron Shapley is more effective in identifying important filters compared to common approaches based on activation patterns. Interestingly, removing just 30 filters with the highest Shapley scores effectively destroys the prediction accuracy of Inception-v3 on ImageNet. Visualization of these few critical filters provides insights into how the network functions. Neuron Shapley is a flexible framework and can be applied to identify responsible neurons in many tasks. We illustrate additional applications of identifying filters that are responsible for biased prediction in facial recognition and filters that are vulnerable to adversarial attacks. Removing these filters is a quick way to repair models. Enabling all these applications is a new multi-arm bandit algorithm that we developed to efficiently estimate Neuron Shapley values.