Smirnakis Lyman Laboratory of Physics Harvard University Cambridge, MA 02138 Lei Xu * Dept. of Computer Science HSH ENG BLDG, Room 1006 The Chinese University of Hong Kong Shatin, NT Hong Kong Abstract Recent work by Becker and Hinton (Becker and Hinton, 1992) shows a promising mechanism, based on maximizing mutual information assumingspatial coherence, by which a system can selforganize itself to learn visual abilities such as binocular stereo. We introduce a more general criterion, based on Bayesian probability theory, and thereby demonstrate a connection to Bayesian theories ofvisual perception and to other organization principles for early vision (Atick and Redlich, 1990). Methods for implementation usingvariants of stochastic learning are described and, for the special case of linear filtering, we derive an analytic expression for the output. 1 Introduction The input intensity patterns received by the human visual system are typically complicated functions of the object surfaces and light sources in the world. It *Lei Xu was a research scholar in the Division of Applied Sciences at Harvard University while this work was performed. Thus the visual system must be able to extract information from the input intensities that is relatively independent of the actual intensity values.

Padhraic Smyth Jet Propulsion Laboratory 238-420 California Institute of Technology 4800 Oak Grove Drive Pasadena, CA 91109 Abstract This paper describes probabilistic methods for novelty detection when using pattern recognition methods for fault monitoring of dynamic systems. The problem of novelty detection is particularly acutewhen prior knowledge and training data only allow one to construct an incomplete classification model. Allowance must be made in model design so that the classifier will be robust to data generated by classes not included in the training phase. For diagnosis applications one practical approach is to construct both an input density model and a discriminative class model. Using Bayes' rule and prior estimates of the relative likelihood of data of known and unknown origin the resulting classification equations are straightforward.

In solving these tasks, one is faced with a large variety of learning algorithms and a vast selection of possible network architectures. After all the training, how does one know which is the best network? This decision is further complicated by the fact that standard techniques can be severely limited by problems such as over-fitting, data sparsity and local optima. The usual solution to these problems is a winner-take-all cross-validatory model selection. However, recent experimental and theoretical work indicates that we can improve performance by considering methods for combining neural networks.

The conventional Bayesian justification of backprop is that it finds the MAP weight vector. As this paper shows, to find the MAP io function instead one must add a correction tenn to backprop. That tenn biases one towards io functions with small description lengths, and in particular favors (somekinds of) feature-selection, pruning, and weight-sharing.

Signal processing and classification algorithms often have limited applicability resulting from an inaccurate model of the signal's underlying structure.We present here an efficient, Bayesian algorithm for modeling a signal composed of the superposition of brief, Poisson-distributed functions. This methodology is applied to the specific problem of modeling and classifying extracellular neural waveforms which are composed of a superposition of an unknown number of action potentials CAPs). Previous approaches have had limited success due largely to the problems of determining the spike shapes, deciding how many are shapes distinct, and decomposing overlapping APs. A Bayesian solution to each of these problems is obtained by inferring a probabilistic model of the waveform. This approach quantifies the uncertainty of the form and number of the inferred AP shapes and is used to obtain an efficient method for decomposing complex overlaps. This algorithm can extract many times more information than previous methods and facilitates the extracellular investigation of neuronal classes and of interactions within neuronal circuits.

To assess the state of the art and identify issues requiring further investigation, a workshop on qualitative and abstract probability was held during the third week of November 1993. This workshop brought together a mix of active researchers from academia, industry, and government interested in the practical and theoretical impact of these abstractions on techniques, methods, and tools for solving complex AI tasks. The result was a set of specific recommendations on the most promising and important avenues for future research.

This paper is a multidisciplinary review of empirical, statistical learning from a graphical model perspective. Well-known examples of graphical models include Bayesian networks, directed graphs representing a Markov chain, and undirected networks representing a Markov field. These graphical models are extended to model data analysis and empirical learning using the notation of plates. Graphical operations for simplifying and manipulating a problem are provided including decomposition, differentiation, andthe manipulation of probability models from the exponential family. Two standard algorithm schemas for learning are reviewed in a graphical framework: Gibbs sampling and the expectation maximizationalgorithm. Using these operations and schemas, some popular algorithms can be synthesized from their graphical specification. This includes versions of linear regression, techniques for feed-forward networks, and learning Gaussian and discrete Bayesian networks from data. The paper concludes by sketching some implications for data analysis and summarizing how some popular algorithms fall within the framework presented. The main original contributions here are the decompositiontechniques and the demonstration that graphical models provide a framework for understanding and developing complex learning algorithms.

The attempt to find a single "optimal" weight vector in conventional network training can lead to overfitting and poor generalization. Bayesian methods avoid this, without the need for a validation set, by averaging the outputs of many networks with weights sampled from the posterior distribution given the training data. This sample can be obtained by simulating a stochastic dynamical system that has the posterior as its stationary distribution.

We present a methodological framework enabling a detailed description of the performance of Hopfield-like attractor neural networks (ANN) in the first two iterations. Using the Bayesian approach, we find that performance is improved when a history-based term is included in the neuron's dynamics. A further enhancement of the network's performance is achieved by judiciously choosing the censored neurons (those which become active in a given iteration) on the basis of the magnitude of their post-synaptic potentials. The contribution of biologically plausible, censored, historydependent dynamics is especially marked in conditions of low firing activity and sparse connectivity, two important characteristics of the mammalian cortex. In such networks, the performance attained is higher than the performance of two'independent' iterations, which represents an upper bound on the performance of history-independent networks.

We analyze the "query by committee" algorithm, a method for filtering informative queries from a random stream of inputs. We show that if the two-member committee algorithm achieves information gain with positive lower bound, then the prediction error decreases exponentially with the number of queries. We show that, in particular, this exponential decrease holds for query learning of thresholded smooth functions.