In the last few years, interest has grown in exploring AI approaches to design problems, both because of the enormous potential impact on productivity of improved design tools and because of the interesting basic AI issues that these problems raise. In particular, a number of ship designers and AI researchers recently became interested in applying AI to the hydrodynamic design of ship hulls. A typical problem here is to design the shape of a ship's hull in response to desired hydrodynamic properties such as drag and stability, taking into consideration a variety of design constraints, such as total hull volume.

In ICML-90, pp. 216–224.

A full book, available for free in PDF form.From the preface:A major problem in artificial intelligence is to endow computers with commonsense knowledge of the world and with the ability to use that knowledge sensibly. A large body of research has studied this problem through careful analysis of typical examples of reasoning in a variety of commonsense domains. The immediate aim of this research is to develop a rich language for expressing commonsense knowledge, and inference techniques for carrying out commonsense reasoning. This book provides an introduction and a survey of this body of research. It is, to the best of my knowledge, the first book to attempt this.The book is designed to be used as a textbook for a one-semester graduate course on knowledge representation.Morgan Kaufmann

Yoshihiro Morl Kazuhiko Yokosawa ATR Auditory and Visual Perception Research Laboratories 2-1-61 Shiromi Higashiku Osaka 540 Japan ABSTRACT A neural network is applied to the problem of recognizing Kanji characters. The recognition accuracy was higher than that of conventional methods. An analysis of connection weights showed that trained networks can discern the hierarchical structure of Kanji characters. This strategy of trained networks makes high recognition accuracy possible. Our results suggest that neural networks are very effective for Kanji character recognition. 1 INTRODUCTION Neural networks are applied to recognition tasks in many fields.

The performance of these networks has been better than that of conventional methods.

The performance of these networks has been better than that of conventional methods.

We consider a 2-layer, 3-node, n-input neural network whose nodes compute linear threshold functions of their inputs. We show that it is NPcomplete to decide whether there exist weights and thresholds for the three nodes of this network so that it will produce output consistent with a given set of training examples. We extend the result to other simple networks. This result suggests that those looking for perfect training algorithms cannot escape inherent computational difficulties just by considering only simple or very regular networks. It also suggests the importance, given a training problem, of finding an appropriate network and input encoding for that problem. It is left as an open problem to extend our result to nodes with nonlinear functions such as sigmoids.

We present a new hypothesis that the cerebellum plays a key role in actively controlling the acquisition of sensory infonnation by the nervous system. In this paper we explore this idea by examining the function of a simple cerebellar-related behavior, the vestibula-ocular reflex or VOR, in which eye movements are generated to minimize image slip on the retina during rapid head movements. Considering this system from the point of view of statistical estimation theory, our results suggest that the transfer function of the VOR, often regarded as a static or slowly modifiable feature of the system, should actually be continuously and rapidly changed during head movements. We further suggest that these changes are under the direct control of the cerebellar cortex and propose experiments to test this hypothesis.

This study evaluates the performance of the multilayer-perceptron and the frequency-sensitive competitive learning network in identifying five commercial aircraft from radar backscatter measurements. The performance of the neural network classifiers is compared with that of the nearest-neighbor and maximum-likelihood classifiers. Our results indicate that for this problem, the neural network classifiers are relatively insensitive to changes in the network topology, and to the noise level in the training data. While, for this problem, the traditional algorithms outperform these simple neural classifiers, we feel that neural networks show the potential for improved performance.

Encouraged by these results we wanted to explore the question, how we might expand on these models to make them useful for the design of speech recognition systems. A problem that emerges as we attempt to apply neural network models to the full speech recognition problem is the problem of scaling. Simply extending neural networks to ever larger structures and retraining them as one monolithic net quickly exceeds the capabilities of the fastest and largest supercomputers. The search complexity of finding a good solutions in a huge space of possible network configurations also soon assumes unmanageable proportions. Moreover, having to decide on all possible classes for recognition ahead of time as well as collecting sufficient data to train such a large monolithic network is impractical to say the least. In an effort to extend our models from small recognition tasks to large scale speech recognition systems, we must therefore explore modularity and incremental learning as design strategies to break up a large learning task into smaller subtasks. Breaking up a large task into subtasks to be tackled by individual black boxes interconnected in ad hoc arrangements, on the other hand, would mean to abandon one of the most attractive aspects of connectionism: the ability to perform complex constraint satisfaction in a massively parallel and interconnected fashion, in view of an overall optimal perfonnance goal.