Y. Reich (Tel-Aviv University) proposed The paper by M. Benaroch (Syracuse University) suggested the use of knowledge-based tools for mass customization of service products; it dealt in general Grundstein (Framatome, France) reported than the other methods. At the macroeconomic and J. Zahavi (both of Tel-Aviv University) level, Deinichenko et al. presented found that genetic algorithms an expert system that utilizes performed even better than a fuzzy knowledge to analyze economic on Artificial Intelligence linear programming model on their Thus, their conclusion was Academy of Sciences) and T. Szapiro (AIEM4) was held in Tel-Aviv, that AI techniques might provide (Warsaw School of Economics) noted Israel, from 8 to 10 January 1996, better results than rigid analytic the lack of models appropriate to the with participants from 13 countries. Service to customers in the financial for discerning patterns in the economic As a matter of course, almost every area was another focus of the and demographic data of developing presentation at the workshop workshop. Lange et al. described a economies. The paper by touched on AI techniques in one way system for customizing investment Edmonds and S. Moss (Manchester or another.

Although RoboCup's primary objective is a Although RoboCup's final target is a world For the Although it is obvious that building a robot robot (physical robot and software agent) to to play a soccer game is an immense challenge, play a soccer game reasonably well, a wide readers might wonder why we propose range of technologies need to be integrated, RoboCup. It is our intention to use RoboCup and numbers of technical breakthroughs as a vehicle to revitalize AI research by offering must be accomplished. When the accomplishment intelligent robotics, and sensor fusion. of such a goal has significant social The first RoboCup, RoboCup-97, will be held impact, it is considered a grand-challenge during the Fifteenth International Joint Conference project (Kitano et al. 1993). Building a robot on Artificial Intelligence (IJCAI-97) in to play a soccer game itself does not generate Nagoya, Japan, as part of IJCAI-97's special significant social and economic impact, but Accessibility Complete Incomplete and architectures can be evaluated. Computer Sensor Readings Symbolic Nonsymbolic chess is a typical example of the standard Control Central Distributed problem. Various search algorithms were evaluated and developed using this domain. With the recent accomplishment by the Deep Blue team, which beat Kasparov, a human grand Table 1. A major reason for the success of computer chess as a standard problem is that the evaluation of the accomplishment will certainly be considered the progress was clearly defined.

Over the last few years the Fourier Transform (FT) representation of boolean functions has been an instrumental tool in the computational learning theory community. It has been used mainly to demonstrate the learnability of various classes of functions with respect to the uniform distribution.

By applying a likelihood splitting criteria to each expert in the HME we "grow" the tree adaptively during training. Secondly, by considering only the most probable path through the tree we may "prune" branches away, either temporarily, or permanently if they become redundant. We demonstrate results for the growing and path pruning algorithms which show significant speed ups and more efficient use of parameters over the standard fixed structure in discriminating between two interlocking spirals and classifying 8-bit parity patterns. INTRODUCTION The HME (Jordan & Jacobs 1994) is a tree structured network whose terminal nodes are simple function approximators in the case of regression or classifiers in the case of classification. The outputs of the terminal nodes or experts are recursively combined upwards towards the root node, to form the overall output of the network, by "gates" which are situated at the non-terminal nodes.

Over the last few years the Fourier Transform (FT) representation of boolean functions has been an instrumental tool in the computational learning theory community. It has been used mainly to demonstrate the learnability of various classes of functions with respect to the uniform distribution.

We propose a new learning method, "Generalized Learning Vector Quantization (GLVQ)," in which reference vectors are updated based on the steepest descent method in order to minimize the cost function. The cost function is determined so that the obtained learning rule satisfies the convergence condition. We prove that Kohonen's rule as used in LVQ does not satisfy the convergence condition and thus degrades recognition ability. Experimental results for printed Chinese character recognition reveal that GLVQ is superior to LVQ in recognition ability.

In this paper we examine a perceptron learning task. The task is realizable since it is provided by another perceptron with identical architecture. Both perceptrons have nonlinear sigmoid output functions. The gain of the output function determines the level of nonlinearity of the learning task. It is observed that a high level of nonlinearity leads to overfitting. We give an explanation for this rather surprising observation and develop a method to avoid the overfitting. This method has two possible interpretations, one is learning with noise, the other cross-validated early stopping.

A Bayesian-Kullback learning scheme, called Ying-Yang Machine, is proposed based on the two complement but equivalent Bayesian representations for joint density and their Kullback divergence. Not only the scheme unifies existing major supervised and unsupervised learnings, including the classical maximum likelihood or least square learning, the maximum information preservation, the EM & em algorithm and information geometry, the recent popular Helmholtz machine, as well as other learning methods with new variants and new results; but also the scheme provides a number of new learning models. 1 INTRODUCTION Many different learning models have been developed in the literature. We may come to an age of searching a unified scheme for them. With a unified scheme, we may understand deeply the existing models and their relationships, which may cause cross-fertilization on them to obtain new results and variants; We may also be guided to develop new learning models, after we get better understanding on which cases we have already studied or missed, which deserve to be further explored. Recently, a Baysian-Kullback scheme, called the YING-YANG Machine, has been proposed as such an effort(Xu, 1995a). It bases on the Kullback divergence and two complement but equivalent Baysian representations for the joint distribution of the input space and the representation space, instead of merely using Kullback divergence for matching un-structuralized joint densities in information geometry type learnings (Amari, 1995a&b; Byrne, 1992; Csiszar, 1975).

Search for the largest (or the smallest) among a number of input data, Le., the winner-take-all (WTA) action, is an essential part of intelligent data processing such as data retrieval in associative memories [3], vector quantization circuits [4], Kohonen's self-organizing maps [5] etc. In addition to the maximum or minimum search, data sorting also plays an essential role in a number of signal processing such as median filtering in image processing, evolutionary algorithms in optimizing problems [6] and so forth.

We propose an active learning method with hidden-unit reduction.