Genetic Synthesis of Task-Oriented Neural Networks
A stochastic search technique based on genetic algorithms for design of task-oriented neural networks is described in the paper. Although the theory of the algorithms is clear, its implementation in the design of neural structures is not yet well investigated. With the help of two case studies, we want to outline the new design approach.
KeywordsGenetic Algorithm Genetic Operator Excitatory Input Inhibitory Input Evolution Procedure
Unable to display preview. Download preview PDF.
- 1.Montana, D.J., Davis, L.: Training feedforward neural networks using genetic algorithms. Proc. of the 11-th Int. Joint Conf. on AI, pp 762–766, 1989Google Scholar
- 2.Whitley, D., Hanson, T.: Optimising neural networks using faster more accurate genetic search. In J.D. Schaffer (ed.), Proceedings of the Third Int. Conf. on Genetic Algorithms, 1989Google Scholar
- 3.Harp, S.A., Samad, T.: Genetic Synthesis of Neural Network Architecture. In L. Davis (ed.), Handbook of genetic algorithms, 1991Google Scholar
- 4.Cliff, D., Harvey, I., Husbands, P.: Visual Sensory-Motor Networks Without Design: Evolving Visually Guided Robots. Int workshop on Mechatronic Computer Systems for Perception and Action, Halmstad, Sweden, 1992Google Scholar
- 5.Dobnikar, A, Likar, A., Trebar, M.: Real time tracking with Stochastic Reinforcement Learning Algorithm, Int. workshop on Mechatronic Computer Systems for Perception and Action, Halmstad, Sweden, 1992Google Scholar
- 6.Tadel, M., Grabensek L., Dobnikar A.: Neural Networks without design — evolution with genetic algorithms and genotypes of variable length, int. report (in Slovenian language), LASPP-FER, Ljubljana, 1993.Google Scholar