Model-based evolutionary computing: a neural network and genetic algorithm architecture

Traditional evolutionary computing uses an explicit fitness function-mathematical or simulated-to derive a solution to a problem from a population of individuals, over a number of generations. In this paper an architecture is presented which allows such techniques to be used on problems which cannot be expressed mathematically or which are difficult to simulate. A neural network is trained using example individuals with their explicit fitness and the resulting model of the fitness function is then used by the evolutionary algorithm to find a solution. It is shown that the approach is effective over a wide range of function types in comparison to the traditional approach. Finally its application to a user-agent task is described-a system in which the fitness function is purely subjective.