Learning with imperfections - a multi-agent neural-genetic trading system with differing levels of social learning

Some real life dynamic systems are so large and complex that the individuals inside the system can only partially understand their environment. In other words, the dynamic environment is imperfect to its participants. In this paper, by using the stock market as a test bed, we demonstrate an integrated individual learning and social learning model for optimisation problems in dynamic environments with imperfect information. By applying differing levels of social learning process in an evolutionary simulated stock market, we study the importance of social learning on the adaptability of artificial agents in imperfect environments. Comparisons between the integrated individual and social learning model and other evolutionary approaches for dynamic optimisation problems, particularly the memory-based approaches and multipopulation approaches, are also drawn with the emphasis on optimisation problems with imperfect information.