Adapting the Biological SSVEP Response to Artificial Neural Networks

TL;DR

This paper introduces a neuroscience-inspired frequency tagging method to assess neuron importance in ANNs, revealing biologically analogous tuning behaviors and enhancing interpretability and pruning strategies.

Contribution

It presents a novel frequency tagging approach for neuron significance assessment in ANNs, inspired by biological neural responses, improving interpretability and pruning techniques.

Findings

Neuron-specific responses show harmonics and intermodulations under frequency tagging.

ANNs exhibit biological-like tuning to flickering frequencies.

Method enhances understanding of network decision processes.

Abstract

Neuron importance assessment is crucial for understanding the inner workings of artificial neural networks (ANNs) and improving their interpretability and efficiency. This paper introduces a novel approach to neuron significance assessment inspired by frequency tagging, a technique from neuroscience. By applying sinusoidal contrast modulation to image inputs and analyzing resulting neuron activations, this method enables fine-grained analysis of a network's decision-making processes. Experiments conducted with a convolutional neural network for image classification reveal notable harmonics and intermodulations in neuron-specific responses under part-based frequency tagging. These findings suggest that ANNs exhibit behavior akin to biological brains in tuning to flickering frequencies, thereby opening avenues for neuron/filter importance assessment through frequency tagging. The proposed method holds promise for applications in network pruning, and model interpretability, contributing to the advancement of explainable artificial intelligence and addressing the lack of transparency in neural networks. Future research directions include developing novel loss functions to encourage biologically plausible behavior in ANNs.