A central problem in connectionist modelling is the control of network and architectural resources during learning. In the present approach, weights reflect a coarse prediction history as coded by a distribution of values and parameterized in the mean and standard deviation of these weight distributions. Weight updates are a function of both the mean and standard deviation of each connection in the network and vary as a function of the error signal ("stochastic delta rule"; Hanson, 1990). Consequently, the weights maintain information on their central tendency and their "uncertainty" in prediction. Such information is useful in establishing a policy concerning the size of the nodal complexity of the network and growth of new nodes. For example, during problem solving the present network can undergo "meiosis", producing two nodes where there was one "overtaxed" node as measured by its coefficient of variation. It is shown in a number of benchmark problems that meiosis networks can find minimal architectures, reduce computational complexity, and overall increase the efficiency of the representation learning interaction.

A method for storing analog vectors in Hopfield's continuous feedback model is proposed. By analog vectors we mean vectors whose components are real-valued. The vectors to be stored are set as equilibria of the network. The network model consists of one layer of visible neurons and one layer of hidden neurons. We propose a learning algorithm, which results in adjusting the positions of the equilibria, as well as guaranteeing their stability.

We present a novel, modular, recurrent connectionist network architecture which learns to robustly perform incremental parsing of complex sentences. From sequential input, one word at a time, our networks learn to do semantic role assignment, noun phrase attachment, and clause structure recognition for sentences with passive constructions and center embedded clauses. The networks make syntactic and semantic predictions at every point in time, and previous predictions are revised as expectations are affirmed or violated with the arrival of new information. Our networks induce their own "grammar rules" for dynamically transforming an input sequence of words into a syntactic/semantic interpretation. These networks generalize and display tolerance to input which has been corrupted in ways common in spoken language.

In this paper, we discuss a current attempt at applying the organizational principle Edelman calls Neuronal Group Selection to the control of a real, two-link robotic manipulator. We begin by motivating the need for an alternative to the position-control paradigm of classical robotics, and suggest that a possible avenue is to look at the primitive animal limb'neurologically ballistic' control mode. We have been considering a selectionist approach to coordinating a simple perception-action task. 1 MOTIVATION The majority of industrial robots in the world are mechanical manipUlators - often armlike devices consisting of some number of rigid links with actuators mounted where the links join that move adjacent links relative to each other, rotationally or translation ally. At the joints there are typically also sensors measuring the relative position of adjacent links, and it is in terms of position that manipulators are generally controlled (a desired motion is specified as a desired position of the end effector, from which can be derived the necessary positions of the links comprising the manipulator). Position control dominates largely for historical reasons, rooted in bang-bang control: manipulators bumped between mechanical stops placed so as to enforce a desired trajectory for the end effector.

This paper describes a neural network algorithm that (1) performs temporal pattern matching in real-time, (2) is trained online, with a single pass, (3) requires only a single template for training of each representative class, (4) is continuously adaptable to changes in background noise, (5) deals with transient signals having low signalto-noise ratios, (6) works in the presence of non-Gaussian noise, (7) makes use of context dependencies and (8) outputs Bayesian probability estimates. The algorithm has been adapted to the problem of passive sonar signal detection and classification. It runs on a Connection Machine and correctly classifies, within 500 ms of onset, signals embedded in noise and subject to considerable uncertainty. 1 INTRODUCTION This paper describes a neural network algorithm, STOCHASM, that was developed for the purpose of real-time signal detection and classification. Of prime concern was capability for dealing with transient signals having low signal-to-noise ratios (SNR). The algorithm was first developed in 1986 for real-time fault detection and diagnosis of malfunctions in ship gas turbine propulsion systems (Malkoff, 1987). It subsequently was adapted for passive sonar signal detection and classification. Recently, versions for information fusion and radar classification have been developed.

The midbrain of the barn owl contains a map-like representation of sound source direction which is used to precisely orient the head toward targets of interest. Elevation is computed from the interaural difference in sound level. We present models and computer simulations of two stages of level difference processing which qualitatively agree with known anatomy and physiology, and make several striking predictions. 1 INTRODUCTION

The vestibulo-ocular reflex (VOR) is the primary mechanism that controls the compensatory eye movements that stabilize retinal images during rapid head motion. The primary pathways of this system are feed-forward, with inputs from the semicircular canals and outputs to the oculomotor system. Since visual feedback is not used directly in the VOR computation, the system must exploit motor learning to perform correctly. Lisberger(1988) has proposed a model for adapting the VOR gain using image-slip information from the retina. We have designed and tested analog very largescale integrated (VLSI) circuitry that implements a simplified version of Lisberger's adaptive VOR model.

We present an application of back-propagation networks to handwritten digit recognition. Minimal preprocessing of the data was required, but architecture of the network was highly constrained and specifically designed for the task. The input of the network consists of normalized images of isolated digits. The method has 1 % error rate and about a 9% reject rate on zipcode digits provided by the U.S. Postal Service. 1 INTRODUCTION The main point of this paper is to show that large back-propagation (BP) networks can be applied to real image-recognition problems without a large, complex preprocessing stage requiring detailed engineering. Unlike most previous work on the subject (Denker et al., 1989), the learning network is directly fed with images, rather than feature vectors, thus demonstrating the ability of BP networks to deal with large amounts of low level information. Previous work performed on simple digit images (Le Cun, 1989) showed that the architecture of the network strongly influences the network's generalization ability. Good generalization can only be obtained by designing a network architecture that contains a certain amount of a priori knowledge about the problem. The basic design principle is to minimize the number of free parameters that must be determined by the learning algorithm, without overly reducing the computational power of the network.

Fernando J. Pineda Center for Microelectronics Technology Jet Propulsion Laboratory California Institute of Technology Pasadena, CA 91109 ABSTRACT A comparison of algorithms that minimize error functions to train the trajectories of recurrent networks, reveals how complexity is traded off for causality. These algorithms are also related to time-independent fonnalisms. It is suggested that causal and scalable algorithms are possible when the activation dynamics of adaptive neurons is fast compared to the behavior to be learned. Standard continuous-time recurrent backpropagation is used in an example. 1 INTRODUCTION Training the time dependent behavior of a neural network model involves the minimization of a function that measures the difference between an actual trajectory and a desired trajectory. The standard method of accomplishing this minimization is to calculate the gradient of an error function with respect to the weights of the system and then to use the gradient in a minimization algorithm (e.g.

Minimization of energy or error functions has proved to be a useful principle in the design and analysis of neural networks and neural algorithms. A brief list of examples include: the back-propagation algorithm, the use of optimization methods in computational vision, the application of analog networks to the approximate solution of NP complete problems and the Hopfield model of associative memory.