Kickback cuts Backprop's red-tape: Biologically plausible credit assignment in neural networks

TL;DR

This paper introduces Kickback, a biologically plausible and simpler alternative to backpropagation for neural networks, with theoretical guarantees and comparable performance on regression tasks.

Contribution

The paper decomposes backprop into interacting algorithms, derives a new credit assignment method called Kickback, and proves its effectiveness and biological plausibility.

Findings

Kickback matches backprop's performance on real-world regression benchmarks.

The authors provide regret bounds for the sub-algorithms of backprop.

Kickback is significantly simpler than traditional backpropagation.

Abstract

Error backpropagation is an extremely effective algorithm for assigning credit in artificial neural networks. However, weight updates under Backprop depend on lengthy recursive computations and require separate output and error messages -- features not shared by biological neurons, that are perhaps unnecessary. In this paper, we revisit Backprop and the credit assignment problem. We first decompose Backprop into a collection of interacting learning algorithms; provide regret bounds on the performance of these sub-algorithms; and factorize Backprop's error signals. Using these results, we derive a new credit assignment algorithm for nonparametric regression, Kickback, that is significantly simpler than Backprop. Finally, we provide a sufficient condition for Kickback to follow error gradients, and show that Kickback matches Backprop's performance on real-world regression benchmarks.