Intelligent Mutation Rate Control in an Economic Application of Genetic Algorithms

Abstract

This work represents the first application of two-level learning in genetic algorithms in an economic environment in which the fitness value of potential rules are complementary across individuals. Two-level learning, or self-adaptation, incorporates certain strategy parameters into the representation of each individual. In this work, these strategy parameters provide the likelihood of mutation for the individual. These strategy parameters evolve by means of mutation and recombination, just as the object variables do. It is argued that self-adaptation over the parameter governing mutation can replace the election operator proposed by Arifovic (1994) in order to attain convergence to the rational expectations equilibrium. While both adaptive mutation and the election operator are sufficient for convergence, self-adaptation may be more appropriate when being compared with real-world or experimental economic data. Through analysis of a static environment it is shown that this convergence, however, will require a strong selective pressure only attained through a transformation of the baseline fitness function.