Statistical Mechanics Analysis of the Phase Transition for Population Distribution in Complex Adaptive Systems

WANG Bing-hong,CHEN Kan,YUAN Bao-sheng

DOI: https://doi.org/10.3969/j.issn.1672-3813.2004.04.003

2004-01-01

Abstract:We found that in the evolutionary minority game (EMG) with prize-to-fine ratio, R less than 1, transition behavior from self-segregation phase (into opposing groups) to clustering phase (swarming towards cautious group) are universal. The critical point of phase transition depends on not only prize-to-fine ratio R, but also the total number of agents N and the threshold value d of accumulated wealth for agent bankruptcy. We study a simple model for EMG with only three different gene values for strategy(p=0, 1/2, 1)using a statistical mechanics approach. A theory for the steady-state population distribution of the agents is derived based on an ‘adiabatic approximation’ in which the short time fluctuations in the population distribution are integrated out to obtain the effective equation governing the steady-state distribution. It is surprising that the phase transition critical point as the function of the parameters R, N, and d revealed by simple model with only three p values can be applied very well to phase transition in the general EMG model with continuous p value. The transition is determined by two generic factors: the market impact (of the agents′ own actions) and the short time market inefficiency (arbitrage opportunities) due to the fluctuations in the numbers of agents using the opposite strategies. Large market impact favors ‘extreme’ players who choose fixed opposite strategies, while large market inefficiency favors cautious players. The transition depends on the effective rate of strategy switching. When the number of agent N is small, the market impact is relatively large; this favors the extreme behaviors. Frequent strategy switching, on the other hand, leads to clustering of the cautious agents.