Bridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex

TitleBridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex
Publication TypeCBMM Memo
Year of Publication2016
AuthorsLiao Q., Poggio T.
Abstract

We discuss relations between Residual Networks (ResNet), Recurrent Neural Networks (RNNs) and the primate visual cortex. We begin with the observation that a shallow RNN is exactly equivalent to a very deep ResNet with weight sharing among the layers. A direct implementation of such a RNN, although having orders of magnitude fewer parameters, leads to a performance similar to the corresponding ResNet. We propose 1) a generalization of both RNN and ResNet architectures and 2) the conjecture that a class of moderately deep RNNs is a biologically-plausible model of the ventral stream in visual cortex. We demonstrate the effectiveness of the architectures by testing them on the CIFAR-10 dataset.

URLhttps://cbmm.mit.edu/sites/default/files/publications/CBMM%20Memo%20047_arxiv1604.03640v1.pdf
Citation Key223

CBMM Memo No.: 

47