Bidirectional communication in neural networks moderated by a Hebb-like learning rule

TL;DR

This paper introduces a stochastic Hebb-like learning rule enabling neural networks to learn timing tasks in complex topologies, demonstrating improved convergence and bidirectional communication akin to biological systems.

Contribution

The paper presents a novel Hebb-like learning rule that enhances learning performance and introduces pattern presentation order dependence, facilitating bidirectional neural communication.

Findings

Better convergence behavior compared to previous rules

Dependence of pattern presentation order on network learning

Enables bidirectional communication in neural networks

Abstract

We demonstrate that our recently introduced stochastic Hebb-like learning rule is capable of learning the problem of timing in general network topologies generated by an algorithm of Watts and Strogatz. We compare our results with a learning rule proposed by Bak and Chialvo and obtain not only a significantly better convergence behavior but also a dependence of the presentation order of the patterns to be learned by introduction of an additional degree of freedom which allows the neural network to select the next pattern itself whereas the learning rule of Bak and Chialvo stays uneffected. This dependence offers a bidirectional communication between a neuronal and a behavioural level and hence completes the action-perception-cycle which is a characteristics of any living being with a brain.