The training set quality measures for neural network learning

The actual problem of criteria set development for the training sample quality estimation in the problems of neural network learning is solved. It includes such measures as a set monotonicity, complexness, repetition, relative dimensionality, relative dependence approximation simplicity, relative inconsistency, evenness, class separability and compactness, integrated criteria of training set quality evaluation, sample and feature selection criteria. The using of offered criterions in practice allows to automatize the process of a construction, analysis and comparison of neural models for pattern recognition problem solving.