Difference Target Propagation

TL;DR

This paper introduces difference target propagation, a biologically plausible alternative to back-propagation that computes targets instead of gradients, effectively training deep networks with discrete and stochastic units.

Contribution

It proposes difference target propagation using auto-encoders for credit assignment, extending applicability to stochastic and discrete units, and demonstrates competitive performance.

Findings

Effective training of deep networks with discrete units.

Comparable results to back-propagation on continuous units.

State-of-the-art performance on stochastic networks.

Abstract

Back-propagation has been the workhorse of recent successes of deep learning but it relies on infinitesimal effects (partial derivatives) in order to perform credit assignment. This could become a serious issue as one considers deeper and more non-linear functions, e.g., consider the extreme case of nonlinearity where the relation between parameters and cost is actually discrete. Inspired by the biological implausibility of back-propagation, a few approaches have been proposed in the past that could play a similar credit assignment role. In this spirit, we explore a novel approach to credit assignment in deep networks that we call target propagation. The main idea is to compute targets rather than gradients, at each layer. Like gradients, they are propagated backwards. In a way that is related but different from previously proposed proxies for back-propagation which rely on a backwards network with symmetric weights, target propagation relies on auto-encoders at each layer. Unlike back-propagation, it can be applied even when units exchange stochastic bits rather than real numbers. We show that a linear correction for the imperfectness of the auto-encoders, called difference target propagation, is very effective to make target propagation actually work, leading to results comparable to back-propagation for deep networks with discrete and continuous units and denoising auto-encoders and achieving state of the art for stochastic networks.