Study of classical conditioning in Aplysia through the implementation of computational models of its learning circuit

Abstract

The learning phenomenon can be analysed at various levels, but in this paper we treat a specific paradigm of artificial intelligence, i.e. artificial neural networks (ANNs), whose main virtue is their capacity to seek unified and mutually satisfactory solutions which are relevant to biological and psychological models. Many of the procedures and methods proposed previously have used biological and/or psychological principles, models, and data; here, we focus on models which look for a greater degree of coherence. Therefore we analyse and compare all aspects of the Gluck–Thompson and Hawkins ANN models. A multithread computer model is developed for analysis of these models in order to study simple learning phenomena in a marine invertebrate (Aplysia californica) and to check their applicability to research in psychology and neurobiology. The predictive capacities of the models differs significantly: the Hawkins model provides a better analysis of the behavioural repertory of Aplysia on both the associative and the non-associative learning level. The scope of the ANN modelling technique is broadened by integration with neurobiological and behavioural models of associative learning, allowing enhancement of some architectures and procedures that are currently being used.