A new evolutionary algorithm named “Genetic Network Programming, GNP” has been proposed. GNP represents its solutions as directed graphs, which realizes better expression ability than GA and GP which use string and tree structures, respectively. The aim of developing GNP is to deal with dynamic environments efficiently by using the distinguished expression ability and the inherently equipped functions of the network structure. However, since GNP is based on evolution, the programs cannot be changed until one generation ends. In this paper, we propose the extended algorithm, “GNP with Evolution and Learning” which combines evolution and reinforcement learning in order to adapt to dynamic environments quickly. The tileworld is used as a simulation environment and the results show some advantages of the proposed method.