In this paper, a supervised neural network training technique based on constrained optimization is developed for preserving prior knowledge of an input
In the conventional incremental training of support vector machines, candidates for support vectors tend to be deleted if the separating hyperplane rotates as the training data ar...