Learning to synchronize: How biological agents can couple neural task modules for dealing with the stability-plasticity dilemma.

Tom Verguts,Pieter Verbeke

doi:10.1371/journal.pcbi.1006604

Abstract

We provide a novel computational framework on how biological and artificial agents can learn to flexibly couple and decouple neural task modules for cognitive processing. In this way, they can address the stability-plasticity dilemma. For this purpose, we combine two prominent computational neuroscience principles, namely Binding by Synchrony and Reinforcement Learning. The model learns to synchronize task-relevant modules, while also learning to desynchronize currently task-irrelevant modules. As a result, old (but currently task-irrelevant) information is protected from overwriting (stability) while new information can be learned quickly in currently task-relevant modules (plasticity). We combine learning to synchronize with task modules that learn via one of several classical learning algorithms (Rescorla-Wagner, backpropagation, Boltzmann machines). The resulting combined model is tested on a reversal learning paradigm where it must learn to switch between three different task rules. We demonstrate that our combined model has significant computational advantages over the original network without synchrony, in terms of both stability and plasticity. Importantly, the resulting models’ processing dynamics are also consistent with empirical data and provide empirically testable hypotheses for future MEG/EEG studies.

Highlights

Humans and other primates are remarkably flexible in adapting to constantly changing environments
Artificial and biological agents alike face a critical trade-off between being sufficiently adaptive to acquiring novel information and retaining older information; this is known as the stability-plasticity dilemma
Neural task modules can be flexibly coupled and decoupled depending on the task at hand. Testing this framework will allow us to gain more insight in how biological agents deal with the stability-plasticity dilemma

Summary

Introduction

Humans and other primates are remarkably flexible in adapting to constantly changing environments. They excel at integrating information in the long run to detect regularities in the environment and generalize across contexts. Artificial neural networks (ANN), despite being used as models of the primate brain, experience significant problems in these respects. In ANNs, extracting regularities requires slow, distributed learning, which does not allow strong flexibility. Fast sequential learning of different tasks typically leads to (catastrophic) forgetting of previous information (for an overview see [1]). ANNs are typically unable to find a trade-off between being sufficiently adaptive to novel information (plasticity) and retaining older information (stability), a problem known as the stability-plasticity dilemma

Objectives

Methods

Results

Discussion

Conclusion