An application of reinforcement learning to a linear-quadratic, differential game is presented. The reinforcement learning system uses a recently developed algorithm, the residual gradient form of advantage updating. The game is a Markov Decision Process (MDP) with continuous time, states, and actions, linear dynamics, and a quadratic cost function. The game consists of two players, a missile and a plane; the missile pursues the plane and the plane evades the missile. The reinforcement learning algorithm for optimal control is modified for differential games in order to find the minimax point, rather than the maximum. Simulation results are compared to the optimal solution, demonstrating that the simulated reinforcement learning system converges to the optimal answer. The performance of both the residual gradient and non-residual gradient forms of advantage updating and Q-learning are compared. The results show that advantage updating converges faster than Q-learning in all simulations.

The macaque lateral geniculate nucleus (LGN) exhibits an intricate lamination pattern, which changes midway through the nucleus at a point coincident with small gaps due to the blind spot in the retina. We present a three-dimensional model of morphogenesis in which local cell interactions cause a wave of development of neuronal receptive fieldsto propagate through the nucleus and establish two distinct lamination patterns. We examine the interactions between the wave and the localized singularities due to the gaps, and find that the gaps induce the change in lamination pattern. We explore critical factors which determine general LGN organization.

It is well known that a combination of many different predictors can improve predictions. Inthe neural networks community "ensembles" of neural networks has been investigated by several authors, see for instance [1, 2, 3]. Most often the networks in the ensemble are trained individually and then their predictions are combined. This combination is usually done by majority (in classification) or by simple averaging (inregression), but one can also use a weighted combination of the networks.

We investigate the computational power of a formal model for networks ofspiking neurons, both for the assumption of an unlimited timing precision, and for the case of a limited timing precision. We also prove upper and lower bounds for the number of examples that are needed to train such networks.

Understanding knowledge representations in neural nets has been a difficult problem. Principal components analysis (PCA) of contributions (products of sending activations and connection weights) has yielded valuable insights into knowledge representations, but much of this work has focused on the correlation matrix of contributions. The present work shows that analyzing the variance-covariance matrix of contributions yields more valid insights by taking account of weights.

Deformable models are an attractive approach to recognizing nonrigid objects which have considerable within class variability. However, there are severe search problems associated with fitting the to data. We show that by using neural networks to providemodels better starting points, the search time can be significantly reduced. The method is demonstrated on a character recognition task.

A fundamental feature of cortical architecture is its columnar organization, manifested in the tendency of neurons with similar properties to be organized in columns that run perpendicular to the cortical surface. This organization of the cortex was initially discovered by physiological experiments (Mouncastle, 1957; Hubel and Wiesel, 1962), and subsequently confirmed with the demonstration of histologically defined that axonal projections throughout thecolumns. Tracing experiments have shown tend to be organized in vertically aligned clusters or patches.

A Lexical Semantic and Statistical Approach to Lexical Collocation Extraction for Natural Language Generation

The first 50 years of AI are reviewed, and current controversies outlined. Scientific disputes include disagreements over the best research methodology, including classical AI, connectionism, hybrid systems, and situated and evolutionary robotics. Philosophical disputes concern (for instance) whether computation is necessary and sufficient for mentality, whether representations are essential for intelligence, whether consciousness can be explained objectively, and whether the Cartesian presuppositions of (most) AI should be replaced by a neo-Heideggerian approach. With respect to final verdicts, both juries (scientific and philosophical) are still out.

A common difficulty in diagnosing failures within Pratt & Whitney's F100-PW-100/200 gas turbine engine occurs when a fault in one part of a system -- comprising an engine, an airframe, a test cell, and automated ground engine test set (AGETS) equipment -- is manifested as an out-of-bound parameter elsewhere in the system. However, because the self-diagnostics only test the specified local parameter, it will pass, leaving only the operators' experience and traditional fault-isolation manuals to locate the source of the problem in another part of the system. This article describes a diagnostic tool (that is, AGETS MBR), designed to overcome this problem by isolating failures using an overall system troubleshooting approach. AGETS MBR was developed jointly by personnel at Pratt & Whitney and United Technologies Research Center using an AI tool called the qualitative reasoning system (QRS).