Adaptive Learning by Genetic Algorithms

de Herbert Dawid

This book deals with the learning behavior of boundedly rational agents in economic systems. In particular, the modeling of learning populations by genetic algorithms is studied in detail. After an extensive review and discussion of the existing literature in the first part, a mathematical analysis of the dynamic properties of genetic algorithm learning in the general framework of systems with a state dependent fitness function is provided. It is shown that co-evolutionary economic models typically fall into this class and the usefulness of the analytical results derived is illustrated in several game theoretic and microeconomic models. The mathematical analysis is complemented by extensive simulation analyses. The last part of the book demonstrates how the obtained theory may be used to design the algorithm such that the learning of equilibria of the economic system is facilitated.