Domain-specific internet portals are growing in popularity because they gather content from the Web and organize it for easy access, retrieval and search. For example, www.campsearch.com allows complex queries by age, location, cost and specialty over summer camps. This functionality is not possible with general, Web-wide search engines. Unfortunately these portals are difficult and time-consuming to maintain. This paper advocates the use of machine learning techniques to greatly automate the creation and maintenance of domain-specific Internet portals. We describe new research in reinforcement learning, information extraction and text classification that enables efficient spidering, the identification of informative text segments, and the population of topic hierarchies. Using these techniques, we have built a demonstration system: a portal for computer science research papers. It already contains over 50,000 papers and is publicly available at www.cora.justresearch.com. These techniques are ...

Abstract. For large state-space Markovian Decision Problems Monte-Carlo planning is one of the few viable approaches to find near-optimal solutions. In this paper we introduce a new algorithm, UCT, that applies bandit ideas to guide Monte-Carlo planning. In finite-horizon or discounted MDPs the algorithm is shown to be consistent and finite sample bounds are derived on the estimation error due to sampling. Experimental results show that in several domains, UCT is significantly more efficient than its alternatives. 1

The basic tools of machine learning appear in the inner loop of most reinforcement learning algorithms, typically in the form of Monte Carlo methods or function approximation techniques.

TD-Gammon is a neural network that is able to teach itself to play backgammon solely by playing against itself and learning from the results. Starting from random initial play, TD-Gammon’s selfteaching methodology results in a surprisingly strong program: without lookahead, its positional judgement rivals that of human experts, and when combined with shallow lookahead, it reaches a level of play that surpasses even the best human players. The success of TD-Gammon has also been replicated by several other programmers; at least two other neural net programs also appear to be capable of superhuman play. Previous papers on TD-Gammon have focused on developing a scientific understanding of its reinforcement learning methodology. This paper views machine learning as a tool in a programmer’s toolkit, and considers how it can be combined with other programming techniques to achieve and surpass world-class backgammon play. Particular emphasis is placed on programming shallow-depth search algorithms, and on TD-Gammon’s doubling algorithm, which is described in print here for

We describe and evaluate a system for learning domainspecific control knowledge. In particular, given a planning domain, the goal is to output a control policy that performs well on “long random walk ” problem distributions. The system is based on viewing planning domains as very large Markov decision processes and then applying a recent variant of approximate policy iteration that is bootstrapped with a new technique based on random walks. We evaluate the system on the AIPS-2000 planning domains (among others) and show that often the learned policies perform well on problems drawn from the long–random-walk distribution. In addition, we show that these policies often perform well on the original problem distributions from the domains involved. Our evaluation also uncovers limitations of our current system that point to future challenges.

Abstract not found

The execution order of a block of computer instructions on a pipelined machine can make a difference in its running time by a factor of two or more. In order to achieve the best possible speed, compilers use heuristic schedulers appropriate to each specific architecture implementation. However, these heuristic schedulers are time-consuming and expensive to build. We present empirical results using both rollouts and reinforcement learning to construct heuristics for scheduling basic blocks. In simulation, the rollout scheduler outperformed a commercial scheduler, and the reinforcement learning scheduler performed almost as well as the commercial scheduler.

Production scheduling, the problem of sequentially configuring a factory to meet forecasted demands, is a critical problem throughout the manufacturing industry. The requirement of maintaining product inventories in the face of unpredictable demand and stochastic factory output makes standard scheduling models, such as job-shop, inadequate. Currently applied algorithms, such as simulated annealing and constraint propagation, must employ ad-hoc methods such as frequent replanning to cope with uncertainty. In this paper, we describe a Markov Decision Process (MDP) formulation of production scheduling which captures stochasticity in both production and demands. The solution to this MDP is a value function which can be used to generate optimal scheduling decisions online. A simple example illustrates the theoretical superiority of this approach over replanning-based methods. We then describe an industrial application and two reinforcement learning methods for generating an approximate valu...

Value function approximation is a problem central to reinforcement learning. Many applications of reinforcement learning have relied on neural network function approximators, which are very slow to train and require substantial parameter tweaking to obtain good performance. Other reinforcement learning studies have applied nearest neighbor and CMAC function approximators, but these cannot scale to problems with many features, especially if some features are irrelevant. We describe initial work on a new function approximation method that uses regression trees to represent value functions. A novel aspect of our method is its error criterion, which combines three terms: the supervised training error, a Bellman error term, and an advantage error term. By using this composite error criterion, we are able to combine many of the benefits of fitted value iteration, TD(0), and advantage updating. The new method is compared experimentally to previous work that employed TD() to solve job-shop sch...

amy ¡ moss ¡