ABSTRACT
This chapter describes a collection of learning tasks in order to place the most commonly studied learning tasks into a broad context and to outline a range of tasks that are important despite the fact that they are receiving relatively little attention. It addresses learning tasks; methods for solving these tasks are not addressed at all. The chapter suggests that multi-criterion reinforcement learning tasks the critic signals are vectors consisting of the values of multiple performance measures. A critic capable of providing gradient information about a performance measure is usually called a "teacher". The chapter uses that the framework has not been designed to do justice to unsupervised learning tasks. Tasks with distal teachers involve control in more substantive ways than do other supervised learning tasks. A number of special cases of supervised learning tasks with a proximal teacher have been studied separately.
