Efficient Search of Winning Strategies in Multi-agent Systems on Random Network: Importance of Local Solidarity

Multi-agent systems defined on a network can be used for modelling the competition between companies in terms of market dominance. In view of the enormous size of the search space for winning strategies of initial configuration of resource allocation on network, we focus our search on the subspace defined by special local clustering effects, using the recently developed evolutionary computational algorithm. Strategies that emphasize local solidarity, measured by the formation of clusters in the form of triangles linkage between members of the same company, prove to be effective in winning both the market share with high probability and high speed. The result provides a good guideline to improve the collective competitiveness in a network of agents. The formulation is based on the Ising model in statistical physics and the evolutionary game is based on Monte Carlo simulation. Significance and the application of the algorithm in the context of econophysics and damage spreading in network are discussed.