ABSTRACT

Thus far we have been concerned with supervised learning. In all tasks, we have been provided with a set of data objects x1, . . . ,xN and their associated labels (or targets), t1, . . . , tN . For example, objects consisting of Olympic years and targets corresponding to 100 m winning times; objects consisting of documents and targets consisting of document categories. It is the presence of the targets, tn, that makes the tasks supervised.