Abstract

The paper addresses a central problem in evolutionary biology and cognitive science; evolution of a neural based learning phenotype from a structured genotype. It describes morphogenesis of a neural network-based cognitive system, starting from a single genotype having a modulon control structure. It further shows how such a system, denoted as GALA architecture, growing its own recurrent axon connections, can further develop into various structures capable of learning in different learning modes, such as advice learning, reinforcement learning, and emotion learning. The paper particularly considers the emotion learning systems and their motivational structure. A simulation experiment is provided to illustrate the theoretical issues discussed