Optimisation de reseaux de neurones pour la reconnaissance de chiffres manuscrits isoles: Selection et ponderation des primitives par algorithmes genetiques (French text)

Feature selection is a significant stage in all pattern recognition systems. Feature selection can be considered as a global combinatorial optimization problem and is an active research subject. The first goal through the task of feature selection process is to reduce the number of features by throwing out the redundant and irrelevant features from the set of features. The second objective is also to maintain and/or to increase the performance of the classifier used by the recognition system. The genetic algorithms are used to solve the problem of feature selection for the recognition of the isolated handwritten digits.; The results obtained allow reducing the complexity of the neural networks used. The number of features was decrease by 25% compared to the subset of features that are extracted from the recognition system on the isolated handwritten digits while maintaining the performance.