login
Home / Papers / Reinforcement Learning

Reinforcement Learning

201 Citations2014
Marco A. Wiering, M. V. Otterlo
journal unavailable

The main goal of this book is to present an up-to-date series of survey articles on the main contemporary sub-fields of reinforcement learning, including surveys on partially observable environments, hierarchical task decompositions, relational knowledge representation and predictive state representations.

Abstract

random of physical phenomena. Experiments have documented a relationship between reinforced variability and “voluntary choice” (Neuringer and Jensen 2010). For example, human participants judged the choices of “actors” on a computer screen as approximating a voluntary agent when the responses combined two qualities: “matching” choice proportions to reinforcement proportions, this demonstrating the functional nature of the choices; and varying levels of variability/predictability. Voluntary acts are hypothesized to be functional responses that are more or less predictable depending upon the demands of the situation, i.e., the contingencies of reinforcement. Many questions remain. Research is underway attempting to identify the underlying physiological bases of stochastic-response emission, e.g., when male songbirds generate variable songs and when monkeys choose among alternative options (Glimcher 2005). Operant variability procedures have been applied to individuals with psychological disabilities such as attention deficit hyperactivity disorder, autism, and depression. Drugs have been used to distinguish between memory-based and stochastic emission of variable responses. Studies with animal models have shown that reinforced variability facilitates acquisition of difficult-to-learn responses, but attempts to replicate with humans have not been successful. Other areas of initial studies include the influences of reinforced variability on problem solving, motor skills learning, and artistic creativity (see Neuringer 2002 for a review). Each of these areas provides rich questions for additional research.