Authors: Yingni She, Ho-Fung Leung
Addresses: Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong, China. ' Department of Computer Science and Engineering, The Chinese University of Hong Kong, Hong Kong, China.
Abstract: In the traditional minority game, each agent chooses the highest-score predictor at every time step from its initial predictors which are allocated randomly. In this paper, we study a version of the minority game in which one individual privileged agent is allowed to join the game with different memory size from the other agents and free to choose any predictor, while each of the other agents owns small number of predictors. We investigate the privileged agent|s wealth in different dynamic environments. Simulations show that the privileged agent using the proposed intelligent strategy can outperform the other agents in the same model and other models proposed in previous work in terms of individual wealth. We also discuss the impacts of the parameters on the privileged agent|s wealth, such as the number of predictors the privileged agent owns and its memory size. Moreover, we discuss the impact of the number of predictors the other agents possess and their memory sizes on the privileged agent|s wealth.
Keywords: adaptation; dynamic environments; minority games; utility; privileged agents; predictors; simulation; individual wealth.
International Journal of Autonomous and Adaptive Communications Systems, 2009 Vol.2 No.3, pp.226 - 255
Available online: 26 Jun 2009 *Full-text access for editors Access for subscribers Purchase this article Comment on this article