Reducing neural network training data using support vectors

A simple, distributed and sequential procedure has been implemented that include computation of support vectors and use of support vectors data as input to neural networks training. This way we are able to reduce training data to one third and one sixth of original data set after using support vector machines training with radial basis function and polynomial kernels, respectively. The two times training using support vector machines and neural networks do not alter stand of proposed procedure in real life use in view to see one time cost of support vector machines training, reduction in size of original data set for fast training of neural networks and the classification results achieved. This way we are able to select support vector machines and neural networks in machine learning where one technique is support for other technique.