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.

Compensatory information is available from the visual speech signals around the speaker's mouth. Previous attempts at using these visual speech signals to improve automatic speech recognition systems have combined the acoustic and visual speech information at a symbolic level using heuristic rules. In this paper, we demonstrate an alternative approach to fusing the visual and acoustic speech information by training feedforward neural networks to map the visual signal onto the corresponding short-term spectral amplitude envelope (STSAE) of the acoustic signal. This information can be directly combined with the degraded acoustic STSAE. Significant improvements are demonstrated in vowel recognition from noise-degraded acoustic signals. These results are compared to the performance of humans, as well as other pattern matching and estimation algorithms. 1 INTRODUCTION

We are developing a phoneme based.

Eight neural net and conventional pattern classifiers (Bayesianunimodal Gaussian, k-nearest neighbor, standard back-propagation, adaptive-stepsize back-propagation, hypersphere, feature-map, learning vector quantizer, and binary decision tree) were implemented on a serial computer and compared using two speech recognition and two artificial tasks. Error rates were statistically equivalent on almost all tasks, but classifiers differed by orders of magnitude in memory requirements, training time, classification time, and ease of adaptivity. Nearest-neighbor classifiers trained rapidly but required the most memory. Tree classifiers provided rapid classification but were complex to adapt. Back-propagation classifiers typically required long training times and had intermediate memory requirements. These results suggest that classifier selection should often depend more heavily on practical considerations concerning memory and computation resources, and restrictions on training and classification times than on error rate.

Traditional methods of studying neural coding characterize the encoding of known stimuli in average neural responses. Organisms face nearly the opposite task - decoding short segments of a spike train to extract information about an unknown, time-varying stimulus. Here we present strategies for characterizing the neural code from the point of view of the organism, culminating in algorithms for real-time stimulus reconstruction based on a single sample of the spike train. These methods are applied to the design and analysis of experiments on an identified movement-sensitive neuron in the fly visual system. As far as we know this is the first instance in which a direct "reading" of the neural code has been accomplished.

We have done an empirical study of the relation of the number of parameters (weights) in a feedforward net to generalization performance.

In this paper, we present a novel implementation of the widely used Back-propagation neural net learning algorithm on the Connection Machine CM-2 - a general purpose, massively parallel computer with a hypercube topology. This implementation runs at about 180 million interconnections per second (IPS) on a 64K processor CM-2. The main interprocessor communication operation used is 2D nearest neighbor communication. The techniques developed here can be easily extended to implement other algorithms for layered neural nets on the CM-2, or on other massively parallel computers which have 2D or higher degree connections among their processors. 1 Introduction High-speed simulation of large artificial neural nets has become an important tool for solving real world problems and for studying the dynamic behavior of large populations of interconnected processing elements [3, 2]. This work is intended to provide such a simulation tool for a widely used neural net learning algorithm - the Back-propagation (BP) algorithm.[7] The hardware we have used is the Connection Machine CM-2.2

Oxford OX1 3PJ Edinburgh EH9 3JL U niv. of Edinburgh ABSTRACT We announce new CMOS synapse circuits using only three and four MOSFETsisynapse. Neural states are asynchronous pulse streams, upon which arithmetic is performed directly. Chips implementing over 100 fully programmable synapses are described and projections to networks of hundreds of neurons are made. 1 OVERVIEW OF PULSE FIRING NEURAL VLSI The inspiration for the use of pulse firing in silicon neural networks is clearly the electrical/chemical pulse mechanism in "real" biological neurons. Neurons fire voltage pulses of a frequency determined by their level of activity but of a constant magnitude (usually 5 Volts) [Murray,1989a]. As indicated in Figure 1, synapses perform arithmetic directly on these asynchronous pulses, to increment or decrement the receiving neuron's activity.

We introduce a cost function for learning in feed-forward neural networks which is an explicit function of the internal representation in addition to the weights. The learning problem can then be formulated as two simple perceptrons and a search for internal representations. Back-propagation is recovered as a limit. The frequency of successful solutions is better for this algorithm than for back-propagation when weights and hidden units are updated on the same timescale i.e. once every learning step. 1 INTRODUCTION In their review of back-propagation in layered networks, Rumelhart et al. (1986) describe the learning process in terms of finding good "internal representations" of the input patterns on the hidden units. However, the search for these representations is an indirect one, since the variables which are adjusted in its course are the connection weights, not the activations of the hidden units themselves when specific input patterns are fed into the input layer. Rather, the internal representations are represented implicitly in the connection weight values. More recently, Grossman et al. (1988 and 1989)1 suggested a way in which the search for internal representations could be made much more explicit.

Decision making tasks that involve delayed consequences are very common yet difficult to address with supervised learning methods. If there is an accurate model of the underlying dynamical system, then these tasks can be formulated as sequential decision problems and solved by Dynamic Programming. This paper discusses reinforcement learning in terms of the sequential decision framework and shows how a learning algorithm similar to the one implemented by the Adaptive Critic Element used in the pole-balancer of Barto, Sutton, and Anderson (1983), and further developed by Sutton (1984), fits into this framework. Adaptive neural networks can play significant roles as modules for approximating the functions required for solving sequential decision problems.