Semi-supervised learning with the help of Parzen windows (Q640960)

Learning theory provides mathematical foundations of many algorithms for learning from examples. A large class of learning algorithms are generated by kernel-based regularization schemes. There are two kinds of kernel-based regularization schemes: one for supervised learning and the other for semi-supervised learning or unsupervised learning. In this paper the authors introduce Parzen windows to the classical least squares regularized regression with the motivation of using unlabelled data for semi-supervised learning. The main results are an error analysis and satisfactory learning rates for the learning algorithm.