Multifunctional learning of a multi-agent based evolutionary artificial neural network with lifetime learning

Inspired by multifunctional neural networks in biological brain, this paper is concerned with building multifunctional learning ability for artificial neural networks. A multi-agent based evolutionary artificial neural network with lifetime learning (MENL) is used to learn two kinds of navigation abilities together: to explore unknown environments as far as possible, and to reach designated goals in the environments. Since these two functions share the same network mechanism and the common knowledge about subject behavior decision and environmental information processing, the learning of one function can benefit the learning of another. This concept has been demonstrated by satisfactory experimental results. Detailed discussion has concluded that the strategies of evolutionary multi-agents and lifetime learning used in MENL are beneficial to the successful multifunctional learning of MENL.