Improved stability and convergence with three factor learning

Authors:
Bernd Porr;Tomas Kulvicius;Florentin Wörgötter
Affiliations:
Department of Electronics & Electrical Engineering, University of Glasgow, Glasgow, GT12 8LT, UK;Bernstein Center of Computational Neuroscience, University Göttingen, Germany;Bernstein Center of Computational Neuroscience, University Göttingen, Germany and Computational Neuroscience, Psychology, University of Stirling, FK9 4LR Stirling, UK
Venue:
Neurocomputing
Year:
2007

Citing 3
Cited 1

Differential Hebbian learning

AIP Conference Proceedings 151 on Neural Networks for Computing
Introduction to Reinforcement Learning

Introduction to Reinforcement Learning
Isotropic sequence order learning

Neural Computation

Discretized ISO-learning neural network for obstacle avoidance in reactive robot controllers

Neurocomputing

Quantified Score

Hi-index	0.01

Visualization

Abstract

Donald Hebb postulated that if neurons fire together they wire together. However, Hebbian learning is inherently unstable because synaptic weights will self-amplify themselves: the more a synapse drives a postsynaptic cell the more the synaptic weight will grow. We present a new biologically realistic way of showing how to stabilise synaptic weights by introducing a third factor which switches learning on or off so that self-amplification is minimised. The third factor can be identified by the activity of dopaminergic neurons in ventral tegmental area which leads to a new interpretation of the dopamine signal which goes beyond the classical prediction error hypothesis.