Bridging the Gaps Between Residual Learning, Recurrent Neural Networks and Visual Cortex
Name
CBMM-Memo-047.pdf
Size
1.29 MB
Format
Adobe PDF
Checksum (MD5)
4f0cd81fa01acd9a31b86cf857e3311f
Author(s) •
Liao, Qianli
Poggio, Tomaso
Date Issued
April 12, 2016
Publisher
Center for Brains, Minds and Machines (CBMM), arXiv
Citation
arXiv:1604.03640v1
Series/Report no.
CBMM Memo Series;047
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.
Subjects
Residual Networks (ResNet)
Recurrent Neural Networks (RNNs)
primate visual cortex
CIFAR-10
Terms of Use
Attribution-NonCommercial-ShareAlike 3.0 United States
Persistent DSpace Link