New composite evolutionary computation algorithm using interactions among genetic evolution, individual learning and social learning

This paper studies the characteristics of a new composite evolutionary computation algorithm in which genetic evolution, individual learning and social learning interact in NK fitness landscape. We derive conditions for effective social learning in static and dynamic environments using computer simulations of a model of the composite evolutionary algorithm. The conditions for static environments are: the individual learning cost should be at least 1.5 times than the social one; the mutation rate should be less than 0.04 per each gene; more than 3 genes should not interact. These conditions qualitatively mean that: the individual learning cost is larger than the social learning cost; teaching is beneficial for teachers; mutation rate is not too high, must be smaller than error thresold; the fitness landscape is not so complex. We also show that this algorithm is effective in dynamic environments in which NK fitness landscape changes with time, if these conditions are satisfied. Frequent environmental change favors social learning, but under more severe conditions, such as high epistasis and higher mutation rate than the error threshold, individual learning is more useful in finding better solutions.