This paper surveys the field of reinforcement learning from a computer-science perspective. It is written to be accessible to researchers familiar with machine learning. Both the historical basis of the field and a broad selection of current work are summarized. Reinforcement learning is the problem faced by an agent that learns behavior through trial-and-error interactions with a dynamic environment. The work described here has a resemblance to work in psychology, but differs considerably in the details and in the use of the word "reinforcement." The paper discusses central issues of reinforcement learning, including trading off exploration and exploitation, establishing the foundations of the field via Markov decision theory, learning from delayed reinforcement, constructing empirical models to accelerate learning, making use of generalization and hierarchy, and coping with hidden state. It concludes with a survey of some implemented systems and an assessment of the practical utility of current methods for reinforcement learning.

Abstract not found

This paper presents a new approach to hierarchical reinforcement learning based on decomposing the target Markov decision process (MDP) into a hierarchy of smaller MDPs and decomposing the value function of the target MDP into an additive combination of the value functions of the smaller MDPs. The decomposition, known as the MAXQ decomposition, has both a procedural semantics---as a subroutine hierarchy---and a declarative semantics---as a representation of the value function of a hierarchical policy. MAXQ unifies and extends previous work on hierarchical reinforcement learning by Singh, Kaelbling, and Dayan and Hinton. It is based on the assumption that the programmer can identify useful subgoals and define subtasks that achieve these subgoals. By defining such subgoals, the programmer constrains the set of policies that need to be considered during reinforcement learning. The MAXQ value function decomposition can represent the value function of any policy that is consisten...

On large problems, reinforcement learning systems must use parameterized function approximators such as neural networks in order to generalize between similar situations and actions. In these cases there are no strong theoretical results on the accuracy of convergence, and computational results have been mixed. In particular, Boyan and Moore reported at last year's meeting a series of negative results in attempting to apply dynamic programming together with function approximation to simple control problems with continuous state spaces. In this paper, we present positive results for all the control tasks they attempted, and for one that is significantly larger. The most important differences are that we used sparse-coarse-coded function approximators (CMACs) whereas they used mostly global function approximators, and that we learned online whereas they learned offline. Boyan and Moore and others have suggested that the problems they encountered could be solved by using actual outcomes (...

We explore the notion of a tour guide software agent for assisting users browsing the World Wide Web. A Web tour guide agent provides assistance similar to that provided by ahuman tour guide in a museum -- it guides the user along an appropriate path through the collection, based on its knowledge of the user's interests, of the location and relevance of various items in the collection, and of the way in which others have interacted with the collection in the past. This paper describes a simple but operational tour guide, called Web-Watcher, which has given over 5000 tours to people browsing CMU's School of Computer Science Web pages. WebWatcher accompanies users from page to page, suggests appropriate hyperlinks, and learns from experience to improve its advice-giving skills. We describe the learning algorithms used by WebWatcher, experimental results showing their effectiveness, and lessons learned from this case study in Web tour guide agents.

Abstract. This article presents a general class of associative reinforcement learning algorithms for connectionist networks containing stochastic units. These algorithms, called REINFORCE algorithms, are shown to make weight adjustments in a direction that lies along the gradient of expected reinforcement in both immediate-reinforcement tasks and certain limited forms of delayed-reinforcement tasks, and they do this without explicitly computing gradient estimates or even storing information from which such estimates could be computed. Specific examples of such algorithms are presented, some of which bear a close relationship to certain existing algorithms while others are novel but potentially interesting in their own right. Also given are results that show how such algorithms can be naturally integrated with backpropagation. We close with a brief discussion of a number of additional issues surrounding the use of such algorithms, including what is known about their limiting behaviors as well as further considerations that might be used to help develop similar but potentially more powerful reinforcement learning algorithms.

Abstract. To date, reinforcement learning has mostly been studied solving simple learning tasks. Reinforcement learning methods that have been studied so far typically converge slowly. The purpose of this work is thus two-fold: 1) to investigate the utility of reinforcement learning in solving much more complicated learning tasks than previously studied, and 2) to investigate methods that will speed up reinforcement learning. This paper compares eight reinforcement learning frameworks: adaptive heuristic critic (AHC) learning due to Sutton, Q-learning due to Watkins, and three extensions to both basic methods for speeding up learning. The three extensions are experience replay, learning action models for planning, and teaching. The frameworks were investigated using connectionism as an approach to generalization. To evaluate the performance of different frame-works, a dynamic environment was used as a testbed. The enviromaaent is moderately complex and nondetermin-istic. This paper describes these frameworks and algorithms in detail and presents empirical evaluation of the frameworks.

In many classifier systems, the classifier strength parameter serves as a predictor of future payoff and as the classifier's fitness for the genetic algorithm. We investigate a classifier system, XCS, in which each classifier maintains a prediction of expected payoff, but the classifier's fitness is given by a measure of the prediction's accuracy. The system executes the genetic algorithm in niches defined by the match sets, instead of panmictically. These aspects of XCS result in its population tending to form a complete and accurate mapping X x A => P from inputs and actions to payoff predictions. Further, XCS tends to evolve classifiers that are maximally general subject to an accuracy criterion. Besides introducing a new direction for classifier system research, these properties of XCS make it suitable for a wide range of reinforcement learning situations where generalization over states is desirable. Key words Classifier systems, strength, fitness, accuracy, mapping, generalizati...

To appear in: G. Tesauro, D. S. Touretzky and T. K. Leen, eds., Advances in Neural Information Processing Systems 7, MIT Press, Cambridge MA, 1995. A straightforward approach to the curse of dimensionality in reinforcement learning and dynamic programming is to replace the lookup table with a generalizing function approximator such as a neural net. Although this has been successful in the domain of backgammon, there is no guarantee of convergence. In this paper, we show that the combination of dynamic programming and function approximation is not robust, and in even very benign cases, may produce an entirely wrong policy. We then introduce Grow-Support, a new algorithm which is safe from divergence yet can still reap the benefits of successful generalization. 1 INTRODUCTION Reinforcement learning---the problem of getting an agent to learn to act from sparse, delayed rewards---has been advanced by techniques based on dynamic programming (DP). These algorithms compute a value function ...

The standard reinforcement learning view of the involvement of neuromodulatory systems in instrumental conditioning includes a rather straightforward conception of motivation as prediction of sum future reward. Competition between actions is based on the motivating characteristics of their consequent states in this sense. Substantial, careful, experiments reviewed in Dickinson & Balleine, into the neurobiology and psychology of motivation shows that this view is incomplete. In many cases, animals are faced with the choice not between many different actions at a given state, but rather whether a single response is worth executing at all. Evidence suggests that the motivational process underlying this choice has different psychological and neural properties from that underlying action choice. We describe and model these motivational systems, and consider the way they interact.