00951933

j

00951933 Kitano, Hiroaki Neurogenetic learning: An integrated method of designing and training neural networks using genetic algorithms. Physica D 75, No.1-3, 225-238 (1994). 1994 Elsevier Science B.V. (North-Holland), Amsterdam EN neurogenetic learning algorithm neural network

doi:10.1016/0167-2789(94)90285-2

Summary: This paper presents a neurogenetic learning algorithm which is an integrated method of designing and training neural networks using genetic algorithms. The proposed scheme provides an integrated means to design and train neural networks, and use the gradient-descend approach for fine-tuning of the network weights and biases. The salient characteristics of the neurogenetic learning is that designing of the network structure and the weight tuning is performed simultaneously. This is a clear distinction from other combinations of GA and neural network proposed in the past. Experimental results demonstrate that the method provides a magnitude of speed up in convergence than current methods, and exhibits far better scaling property.