Combining Hebbian and reinforcement learning in a minibrain model

Abstract

A toy model of a neural network in which both Hebbian learning and reinforcement learning occur is studied. The problem of `path interference', which makes that the neural net quickly forgets previously learned input-output relations is tackled by adding a Hebbian term (proportional to the learning rate $η$) to the reinforcement term (proportional to $ρ$) in the learning rule. It is shown that the number of learning steps is reduced considerably if $1/4 < η/ρ< 1/2$, i.e., if the Hebbian term is neither too small nor too large compared to the reinforcement term.