In this paper we present TDLeaf(lambda), a variation on the TD(lambda) algorithm that enables it to be used in conjunction with minimax search. We present some experiments in both chess and backgammon which demonstrate its utility and provide comparisons with TD(lambda) and another less radical variant, TD-directed(lambda). In particular, our chess program, ``KnightCap,'' used TDLeaf(lambda) to learn its evaluation function while playing on the Free Internet Chess Server (FICS, fics.onenet.net). It improved from a 1650 rating to a 2100 rating in just 308 games. We discuss some of the reasons for this success and the relationship between our results and Tesauro's results in backgammon.

In this paper we re-investigate windowing for rule learning algorithms. We show that, contrary to previous results for decision tree learning, windowing can in fact achieve significant run-time gains in noise-free domains and explain the different behavior of rule learning algorithms by the fact that they learn each rule independently. The main contribution of this paper is integrative windowing, a new type of algorithm that further exploits this property by integrating good rules into the final theory right after they have been discovered. Thus it avoids re-learning these rules in subsequent iterations of the windowing process. Experimental evidence in a variety of noise-free domains shows that integrative windowing can in fact achieve substantial run-time gains. Furthermore, we discuss the problem of noise in windowing and present an algorithm that is able to achieve run-time gains in a set of experiments in a simple domain with artificial noise.

We study numerically the properties of the bayesian perceptron through a gradient descent on the optimal cost function. The theoretical distribution of stabilities is deduced. It predicts that the optimal generalizer lies close to the boundary of the space of (error-free) solutions. The numerical simulations are in good agreement with the theoretical distribution. The extrapolation of the generalization error to infinite input space size agrees with the theoretical results. Finite size corrections are negative and exhibit two different scaling regimes, depending on the training set size. The variance of the generalization error vanishes for $N \rightarrow \infty$ confirming the property of self-averaging.