Publisher Correction: Limits to visual representational correspondence between convolutional neural networks and the human brain

Abstract

The manuscript by Xu and Vaziri-Pashkam reanalyzes some previously acquired fMRI data in order to assess the accuracy of convolutional neural networks (CNNs) as models of the visual brain. Specifically, the authors take a large number of different CNNs and systematically compare the representations in these networks using RSA (representational similarity analysis) against different regions in human visual cortex. Importantly, the authors have high-quality datasets that were collected in response to a variety of natural and artificial objects.