KenjiDoya ATR International; CREST, JST 2-2 Hikaridai Seika-cho Soraku-gun Kyoto 619-0288 JAPAN doya@isd.atr.co.jp Abstract This paper proposes a new reinforcement learning (RL) paradigm that explicitly takes into account input disturbance as well as modeling errors.The use of environmental models in RL is quite popular for both off-line learning by simulations and for online action planning. However, the difference between the model and the real environment can lead to unpredictable, often unwanted results. Based on the theory of H oocontrol, we consider a differential game in which a'disturbing' agent (disturber) tries to make the worst possible disturbance while a'control' agent (actor) tries to make the best control input. The problem is formulated as finding a minmax solutionof a value function that takes into account the norm of the output deviation and the norm of the disturbance. We derive online learning algorithms for estimating the value function and for calculating the worst disturbance and the best control in reference tothe value function.

It should be stressed that in our model information is coded in the non-stationary rates that underlie the input spike trains. As these rates are not observable, any learning must depends on the observable input spikes that realize those underlying rates.

We present a novel method for clustering using the support vector machine approach.Data points are mapped to a high dimensional feature space, where support vectors are used to define a sphere enclosing them. The boundary of the sphere forms in data space a set of closed contours containing the data. Data points enclosed by each contour are defined as a cluster. As the width parameter of the Gaussian kernel is decreased, these contours fit the data more tightly and splitting of contours occurs. The algorithm works by separating clusters according to valleys in the underlying probabilitydistribution, and thus clusters can take on arbitrary geometrical shapes.

E lRP such as to achieve minimum Bayes error (or probability of misclassification). Two avenues will be explored: the first is to maximize the ()-average divergence between the class densities and the second is to minimize the union Bhattacharyya bound in the range of (). While both approaches yield similar performance in practice, they outperform standardLDA features and show a 10% relative improvement in the word error rate over state-of-the-art cepstral features on a large vocabulary telephony speech recognition task. 1 Introduction Modern speech recognition systems use cepstral features characterizing the short-term spectrum of the speech signal for classifying frames into phonetic classes. These features are augmented with dynamic information from the adjacent frames to capture transient spectral events in the signal.

Algebraic geometry is essential to learning theory. In hierarchical learning machines such as layered neural networks and gaussian mixtures, the asymptotic normality does not hold, since Fisher information matricesare singular. In this paper, the rigorous asymptotic form of the stochastic complexity is clarified based on resolution of singularities and two different problems are studied.

Hidetoshi Nishimori Department of Physics, Tokyo Institute of Technology, Oh-Okayama, Meguro-ku, Tokyo 152-8551, Japan nishi@stat.phys.titech.ac.jp Abstract We introduce stagewise processing in error-correcting codes and image restoration, by extracting information from the former stage and using it selectively to improve the performance of the latter one. Both mean-field analysis using the cavity method and simulations showthat it has the advantage of being robust against uncertainties in hyperparameter estimation. 1 Introduction In error-correcting codes [1] and image restoration [2], the choice of the so-called hyperparameters is an important factor in determining their performances. In error correction, they determine the statistical significance given to the paritychecking termsand the received bits. Similarly in image restoration, they determine the statistical weights given to the prior knowledge and the received data. It was shown, by the use of inequalities, that the choice of the hyperparameters is optimal whenthere is a match between the source and model priors [3].

Several regular programs will return is March 15, 2002.

In this paper, we present algorithms that perform gradient ascent of the average reward in a partially observable Markov decision process (POMDP). These algorithms are based on GPOMDP, an algorithm introduced in a companion paper (Baxter & Bartlett, this volume), which computes biased estimates of the performance gradient in POMDPs. The algorithm's chief advantages are that it uses only one free parameter beta, which has a natural interpretation in terms of bias-variance trade-off, it requires no knowledge of the underlying state, and it can be applied to infinite state, control and observation spaces. We show how the gradient estimates produced by GPOMDP can be used to perform gradient ascent, both with a traditional stochastic-gradient algorithm, and with an algorithm based on conjugate-gradients that utilizes gradient information to bracket maxima in line searches. Experimental results are presented illustrating both the theoretical results of (Baxter & Bartlett, this volume) on a toy problem, and practical aspects of the algorithms on a number of more realistic problems.

Gradient-based approaches to direct policy search in reinforcement learning have received much recent attention as a means to solve problems of partial observability and to avoid some of the problems associated with policy degradation in value-function methods. In this paper we introduce GPOMDP, a simulation-based algorithm for generating a biased estimate of the gradient of the average reward in Partially Observable Markov Decision Processes POMDPs controlled by parameterized stochastic policies. A similar algorithm was proposed by (Kimura et al. 1995). The algorithm's chief advantages are that it requires storage of only twice the number of policy parameters, uses one free beta (which has a natural interpretation in terms of bias-variance trade-off), and requires no knowledge of the underlying state. We prove convergence of GPOMDP, and show how the correct choice of the parameter beta is related to the mixing time of the controlled POMDP. We briefly describe extensions of GPOMDP to controlled Markov chains, continuous state, observation and control spaces, multiple-agents, higher-order derivatives, and a version for training stochastic policies with internal states. In a companion paper (Baxter et al., this volume) we show how the gradient estimates generated by GPOMDP can be used in both a traditional stochastic gradient algorithm and a conjugate-gradient procedure to find local optima of the average reward.

For more details, see Nau et al. 's preconditions can include logical inferences, 's preconditions two methods for traveling from one location can include Horn-clause inferencing, numeric to another: (1) traveling by airplane and (2) computations, and calls to external programs. 's expressive power can be used to create a totally ordered list of subtasks. Suppose domain representations for complex application that all these subtasks are primitive except for domains. For example, the Horn 4. if t is primitive (i.e., there is an operator for t) then clauses can include calls to attached procedures 5. nondeterministically choose an operator o for t We believe the primary 14. endif's higher level of expressivity made it possible to formulate highly expressive domain algorithms in's data structures to make them faster; for example, we found that a simple change to the data structure We intend to make more optimizations in the near future. (Aha and Breslow 1997).