Book chapter
Chapter 69 - Flexible Learning of Optimal Strategies
Modeling and Control of Economic Systems 2001, pp 413-418
2003
Abstract
This chapter presents an agent-based computer simulation of equilibrium in some games that do not have dominant strategy equilibria. The agents learn optimal strategies through a flexible learning algorithm somewhat based on simulated annealing. The model proposed is distinguished from the large literature of such models by adopting ideas both from simulated annealing and from behavioral economics. The relation between the aspiration level and the outcomes thus determines the evolution of the resist, leading to the formation of approximately optimal habits in a range of game situations. It was found that agents adapted predictably to the Nash equilibrium, in most cases, as the population distribution of habits changes. In one case, a hawk-and-dove-like coordination game, however, the maximin solution strongly influences the equilibrium. Based on this and other tests, the model succeeds as a model of flexible learning. The one departure from Nash equilibria, in context, suggests that the model can contribute something to the understanding of the limits of that concept in game theory.
Metrics
5 Record Views
Details
- Title
- Chapter 69 - Flexible Learning of Optimal Strategies
- Creators
- Roger A. McCain - Drexel University
- Publication Details
- Modeling and Control of Economic Systems 2001, pp 413-418
- Publisher
- Elsevier Science Ltd
- Resource Type
- Book chapter
- Language
- English
- Academic Unit
- Economics (School of Economics)
- Other Identifier
- 991021807001104721