Search of optimal neural net structures using genetic algorithms and simulated annealing. Verify with the benchmark PROBEN1.
Francisco Yepes Barrera · INTELIGENCIA ARTIFICIAL · 2007
The paper describes the use of Genetic Algorithms (GA) and Simulated Annealing (SA) for the configuration of neural networks, within the framework of a specific architecture, TSAGANN. The comparison has been carried out on established benchmarks, and described in detail. Statistical analysis of results indicates that SA seems not to be penalized with respect to GA, giving comparable results but requiring a minor number of evaluations.