Evolutionary Optimization of Neural Networks Training Set: Application in the Lymphocytes' Nuclei Classification

F. Moreira and M. Roisenberg (Brazil)


Genetic Algorithm, Neural Network, Pattern Recognition.


Moreover all heuristic involved in the training process of a feed-forward neural network, the choice of a representative sample data in the training set is a very important step. With a small but representative training set, the neural network can correctly distinguish the patterns to be recognized and be able to appropriately generalize. In this paper, a genetic algorithm was developed to optimize the training set, to satisfactorily train a feed-forward neural network with a back propagation learning algorithm. The classification of nuclei of lymphocytes of human peripheral blood was used as an application where we apply the methodology.

Important Links:

Go Back