NeurFlow: Interpreting Neural Networks through Neuron Groups and Functional Interactions

TL;DR

NeurFlow is a novel framework that enhances neural network interpretability by analyzing neuron groups and their functional interactions, providing clearer insights into internal processes and practical applications.

Contribution

This paper introduces NeurFlow, a framework that shifts from individual neuron analysis to neuron group interactions, improving interpretability and reducing computational costs.

Findings

NeurFlow accurately identifies core neuron groups.

It constructs hierarchical circuits of neuron interactions.

Demonstrates utility in image debugging and concept labeling.

Abstract

Understanding the inner workings of neural networks is essential for enhancing model performance and interpretability. Current research predominantly focuses on examining the connection between individual neurons and the model's final predictions. Which suffers from challenges in interpreting the internal workings of the model, particularly when neurons encode multiple unrelated features. In this paper, we propose a novel framework that transitions the focus from analyzing individual neurons to investigating groups of neurons, shifting the emphasis from neuron-output relationships to functional interaction between neurons. Our automated framework, NeurFlow, first identifies core neurons and clusters them into groups based on shared functional relationships, enabling a more coherent and interpretable view of the network's internal processes. This approach facilitates the construction of a hierarchical circuit representing neuron interactions across layers, thus improving interpretability while reducing computational costs. Our extensive empirical studies validate the fidelity of our proposed NeurFlow. Additionally, we showcase its utility in practical applications such as image debugging and automatic concept labeling, thereby highlighting its potential to advance the field of neural network explainability.