Forming Neural Networks through E cient and

This article demonstrates the advantages of a cooperative, coevolutionary search in diicult control problems. The SANE system coevolves a population of neurons that cooperate to form a functioning neural network. In this process, neurons assume diierent but overlapping roles, resulting in a robust encoding of control behavior. SANE is shown to be more eecient, more adaptive, and maintain higher levels of diversity than the more common network-based population approaches. Further empirical studies illustrate the emergent neuron specializations and the diierent roles the neurons assume in the population.