Learning with mislabeled training samples using stochastic approximation | IEEE Journals & Magazine | IEEE Xplore