Global optimisation methods for choosing the connectivity pattern of N-tuple classifiers
2004 IEEE International Joint Conference on Neural Networks (IEEE Cat No 04CH37541) IJCNN-04 · 2004
An experimental study on the use of global optimisation methods, such as Genetic Algorithms, Simulated Annealing and Tabu Search, applied to the problem of choosing the connectivity pattern of the N-tuple classifiers is presented. For example, in an experiment, the use of Tabu Search decreased in 17.27% the mean of the classification errors of the networks. In other experiment, the application of Genetic Algorithms not only decreased in 61% the use of memory, but also the mean of the classification errors obtained were lower than the ones initially achieved without this method.