Evolutionary and Hebbian analysis of hierarchically coupled associative memories

Abstract

Inspired by The Theory of Neuronal Group Selection (TNGS) we have conducted a comparative storage and retrieval analysis of a multi-level or hierarchically coupled associative memory through evolutionary computation and Hebbian learning. The TNGS establishes that memory processes can be described as being organized, functionally, in hierarchical levels, where higher levels coordinate sets of functions of the lower levels. The most basic units in the cortical area of the brain are formed during epigenesis and are called neuronal groups, which are defined as a set of localized tightly coupled neurons constituting what we call our first-level blocks of memories. On the other hand the higher levels are formed during our lives, or ontogeny, through selective strengthening or weakening of the neural connections amongst the neuronal groups. In this sense, this paper describes and compares a method of acquiring the inter- group synapses for the proposed coupled system using both evolutionary computation and Hebbian learning. The results show that evolutionary computation, more specifically genetic algorithms, is more suitable for network acquisition than Hebbian learning because it allows for the emergence of complex behaviours which a repotentially excluded due to the well known crossover effect constraints presented in Hebbian learning. Simulations have been carried out considering a wide range of the system parameters.