For the analysis of the dynamics of populations that evolve strategies for games, it is common practice to assume infinitely large populations. Infinite population models yield predictions of fixed points and their stability properties. However, these models are not suited to anticipate the influence of genetic drift, caused by stochastic sampling in small populations. Instead, we propose Markov models of finite populations for the analysis of genetic drift in games. With these exact models, we can study the stability of evolutionary stable strategies, and measure the influence of genetic drift in the long run. We show that genetic drift can introduce significant differences in the expectations of long term behavior.
|Title of host publication||2004 Genetic and Evolutionary Computation COnference (GECCO 2004)|
|Place of Publication||United States, Seattle, Washington|
|Publication status||Published - 2004|