Exploring NK Fitness Landscapes Using an Imitative Learning Search

The idea that a group of cooperating agents can solve problems more efficiently than when those agents work independently is hardly controversial, despite the little quantitative groundwork to support it. Here we investigate the performance of a group of agents in locating the global maxima of NK fitness landscapes with varying degrees of ruggedness. Cooperation is taken into account through imitative learning and the broadcasting of messages informing on the fitness of each agent. We find a trade-off between the group size and the frequency of imitation: for rugged landscapes, too much imitation or too large a group yield a performance poorer than that of independent agents. By decreasing the diversity of the group, imitative learning may lead to duplication of work and hence to a decrease of its effective size. However, when the parameters are set to optimal values the cooperative group substantially outperforms the independent agents.