Evolutionary design of random iterated neural networks
Abstract
An evolutionary algorithm (EA) to search for the topology and parameters of random iterated neural networks (RINN) that realize random dynamical systems is presented. They are asymptotic geometric approximations of given discrete dynamical systems with compact global attractors. The EA used variable-length genotypes to represent the RINN. Using special properties of the problem, some particular genetic operators were designed to improve the effectivity of the EA to search for the solutions which proved well in the experiments.
Publication Title
Intelligent Engineering Systems Through Artificial Neural Networks
Recommended Citation
Nino, F., Hernandez, G., & Dasgupta, D. (1999). Evolutionary design of random iterated neural networks. Intelligent Engineering Systems Through Artificial Neural Networks, 9, 437-443. Retrieved from https://digitalcommons.memphis.edu/facpubs/2792
