Emergence of Learning Rule in Neural Networks Composed of Locally Defined Learning Modules
Hajime Murao, Shinzo Kitamura · IEEJ Transactions on Electronics Information and Systems · 1997
We propose a method to evolve adaptive behavior of an artificial neural network (ANN). The adaptive behavior emerges from the coordination of learning rules. Each learning rule is defined as a function of local information of a corresponding neuron only and modifies the connective strength between the neuron and its neighbor neurons. The learing rule is exposed to the selective pressure based on the fitness value, which represents the importance in producing the adaptive behavior of the ANN. The learning rules with lower fitness values are replaced by new ones generated by the Genetic Programming techniques. Experimental results demonstrate that the proposed method produces adaptive behavior of single-layered ANNs and two-layered ANNs. This means that efficient learning rule evolves. The learning rules in the two-layered ANN coordinate each other and the macroscopic adaptive behavior of the ANN emerges.