分野別セミナー

In the literature, a wide variety of methods have been proposed to deal with the supervised learning problem. From a certain point of view, these methods can be distinguished into two different approaches. The first is to restrict the class of functions that make predictions, for example by only considering linear functions. The second approach is to give a prior probability to every possible function, where higher probabilities are given to functions that we consider to be more likely. The latter raises the problem of dealing with an infinite set of possible prediction functions. Gaussian processes (GPs) are an answer to this problem. GPs provide a principled, practical, probabilistic approach to learning in kernel machines. They provide a well-founded framework for both learning and model selection. Theoretical and practical developments over the last decade have made GPs a serious competitor for real supervised learning applications. In this talk, we will analyze this tool for both the regression and classification problems.