Examining Representational Similarity in ConvNets and the Primate Visual Cortex

TL;DR

This study compares convolutional neural networks with primate visual cortex data, showing that deeper and better-performing models more closely resemble biological neural representations.

Contribution

It demonstrates that increasing ConvNet depth and performance enhances their similarity to primate IT cortex representations.

Findings

Deeper ConvNets have higher representational similarity to primate IT.

Improved validation performance correlates with increased biological plausibility.

ConvNet features increasingly resemble primate visual cortex with depth.

Abstract

We compare several ConvNets with different depth and regularization techniques with multi-unit macaque IT cortex recordings and assess the impact of the same on representational similarity with the primate visual cortex. We find that with increasing depth and validation performance, ConvNet features are closer to cortical IT representations.