Exploiting Spatio-Temporal Structure with Recurrent Winner-Take-All Networks

TL;DR

This paper introduces a convolutional recurrent neural network with Winner-Take-All dropout designed for high-dimensional unsupervised feature learning in multi-dimensional time series, demonstrating improved object recognition in videos.

Contribution

It presents a scalable reinterpretation of Deep Predictive Coding Networks, extends Winner-Take-All Autoencoders to temporal sequences, and introduces new initialization and regularization techniques for convolutional-recurrent networks.

Findings

Achieved better object recognition results in videos compared to existing methods.

Demonstrated effective unsupervised feature learning in high-dimensional time series.

Provided a novel end-to-end trainable architecture with improved regularization.

Abstract

We propose a convolutional recurrent neural network, with Winner-Take-All dropout for high dimensional unsupervised feature learning in multi-dimensional time series. We apply the proposedmethod for object recognition with temporal context in videos and obtain better results than comparable methods in the literature, including the Deep Predictive Coding Networks previously proposed by Chalasani and Principe.Our contributions can be summarized as a scalable reinterpretation of the Deep Predictive Coding Networks trained end-to-end with backpropagation through time, an extension of the previously proposed Winner-Take-All Autoencoders to sequences in time, and a new technique for initializing and regularizing convolutional-recurrent neural networks.