Integration learning is used to improve the classification of a sample, prediction, and functional approximation.