ALVIS is a reinforcement-based connectionist architecture that learns associative maps in continuous multidimensional environments. The discovered locations of positive and negative reinforcements are recorded in "do be" and "don't be" subnetworks, respectively. The outputs of the subnetworks relevant to the current goal are combined and compared with the current location to produce an error vector. This vector is backpropagated through a motor-perceptual mapping network.

While we are waiting for the ultimate biophysics of cell membranes and synapses to be completed, we may speculate on the shapes of neurons and on the patterns of their connections. Much of this will be significant whatever the outcome of future physiology. Take as an example the isotropy, anisotropy and periodicity of different kinds of neural networks. The very existence of these different types in different parts of the brain (or in different brains) defeats explanation in terms of embryology; the mechanisms of development are able to make one kind of network or another. The reasons for the difference must be in the functions they perform.

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.

ABSTRACT A self-organizing Hopfield network has been developed in the context of Vector Ouantiza -tion, aiming at compression of television images. The metastable states of the spin glass-like network are used as an extra storage resource using the Minimal Overlap learning rule (Krauth and Mezard 1987) to optimize the organization of the attractors. The sel f-organi zi ng scheme that we have devised results in the generation of an adaptive codebook for any qiven TV image. As in many applications they are unknown, the aim of this work is to develop a network capable to learn how to select its attractors. TV image compression using Vector Quantization (V.Q.)(Gray, 1984), a key issue for HOTV transmission, is a typical case, since the non neural algorithms which generate the list of codes (the codebookl are suboptimal.

Reconstructing a surface from sparse sensory data is a well-known problem iIi computer vision. This paper describes an experimental analog VLSI chip for smooth surface interpolation from sparse depth data. An eight-node ID network was designed in 3J.lm CMOS and successfully tested.

K. Y. Goldberg and B. A. Pearlmutter School of Computer Science Carnegie Mellon University Pittsburgh, PA 15213 ABSTRACT Computing the inverse dynamics of a robot ann is an active area of research in the control literature. We hope to learn the inverse dynamics by training a neural network on the measured response of a physical ann. The input to the network is a temporal window of measured positions; output is a vector of torques. We train the network on data measured from the first two joints of the CMU Direct-Drive Arm II as it moves through a randomly-generated sample of "pick-and-place" trajectories. We then test generalization with a new trajectory and compare its output with the torque measured at the physical arm.

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.

Erich J. Smythe 207 Greenmanville Ave, #6 Mystic, CT 06355 ABSTRACT SYREN is a connectionist model that uses temporal information in a speech signal for syllable recognition. It classifies the rates and directions of formant center transitions, and uses an adaptive method to associate transition events with each syllable. The system uses explicit spatial temporal representations through delay lines. SYREN uses implicit parametric temporal representations in formant transition classification through node activation onset, decay, and transition delays in sub-networks analogous to visual motion detector cells. SYREN recognizes 79% of six repetitions of 24 consonant-vowel syllables when tested on unseen data, and recognizes 100% of its training syllables. INTRODUCTION Living organisms exist in a dynamic environment. Problem solving systems, both natural and synthetic, must relate and interpret events that occur over time.

We address the question of when a network can be expected to generalize from m random training examples chosen from some arbitrary probability distribution, assuming that future test examples are drawn from the same distribution. Among our results are the following bounds on appropriate sample vs. network size.

The algorithm can automatically find optimal or nearly optimal architectures necessary to solve known Boolean functions, facilitate the interpretation of the activation of the remaining hidden units and automatically estimate the complexity of architectures appropriate for phonetic labeling problems. The general principle of the algorithm can also be adapted to different tasks: for example, it can be used to eliminate the [0, 0] local minimum of the [-1.