MIME-Version: 1.0 Server: CERN/3.0 Date: Monday, 06-Jan-97 22:41:20 GMT Content-Type: text/html Content-Length: 6223 Last-Modified: Monday, 12-Aug-96 18:54:14 GMT Papers To view a paper, click on the open book image.

Uncertain Reasoning

  1. Refinement of Bayesian Networks by Combining Connectionist and Symbolic Techniques
    Sowmya Ramanchandran
    Ph.D. proposal, Department of Computer Sciences, University of Texas at Austin, 1995.

    Bayesian networks provide a mathematically sound formalism for representing and reasoning with uncertain knowledge and are as such widely used. However, acquiring and capturing knowledge in this framework is difficult. There is a growing interest in formulating techniques for learning Bayesian networks inductively. While the problem of learning a Bayesian network, given complete data, has been explored in some depth, the problem of learning networks with unobserved causes is still open. In this proposal, we view this problem from the perspective of theory revision and present a novel approach which adapts techniques developed for revising theories in symbolic and connectionist representations. Thus, we assume that the learner is given an initial approximate network (usually obtained from a expert). Our technique inductively revises the network to fit the data better. Our proposed system has two components: one component revises the parameters of a Bayesian network of known structure, and the other component revises the structure of the network. The component for parameter revision maps the given Bayesian network into a multi-layer feedforward neural network, with the parameters mapped to weights in the neural network, and uses standard backpropagation techniques to learn the weights. The structure revision component uses qualitative analysis to suggest revisions to the network when it fails to predict the data accurately. The first component has been implemented and we will present results from experiments on real world classification problems which show our technique to be effective. We will also discuss our proposed structure revision algorithm, our plans for experiments to evaluate the system, as well as some extensions to the system.

  2. Revising Bayesian Network Parameters Using Backpropagation
    Sowmya Ramachandran and Raymond J. Mooney
    To appear in the Proceedings of the International Conference on Neural Networks (ICNN-96), Washington, D.C., June 1996, Special Session on Knowledge-Based Artificial Neural Networks.

    The problem of learning Bayesian networks with hidden variables is known to be a hard problem. Even the simpler task of learning just the conditional probabilities on a Bayesian network with hidden variables is hard. In this paper, we present an approach that learns the conditional probabilities on a Bayesian network with hidden variables by transforming it into a multi-layer feedforward neural network (ANN). The conditional probabilities are mapped onto weights in the ANN, which are then learned using standard backpropagation techniques. To avoid the problem of exponentially large ANNs, we focus on Bayesian networks with noisy-or and noisy-and nodes. Experiments on real world classification problems demonstrate the effectiveness of our technique.

    Qualitative Modeling & Diagnosis

  3. Learning Qualitative Models for Systems with Multiple Operating Regions
    Sowmya Ramachandran, Raymond J. Mooney and Benjamin J. Kuipers
    Proceedings of the Eight International Workshop of Qualitative Reasoning about Physical Systems, pp. 212-223, Nara, Japan, June 1994. (QR-94)
    The problem of learning qualitative models of physical systems from observations of its behaviour has been addressed by several researchers in recent years. Most current techniques limit themselves to learning a single qualitative differential equation to model the entire system. However, many systems have several qualitative differential equations underlying them. In this paper, we present an approach to learning the models for such systems. Our technique divides the behaviours into segments, each of which can be explained by a single qualitative differential equation. The qualitative model for each segment can be generated using any of the existing techniques for learning a single model. We show that results of applying our technique to several examples and demonstrate that it is effective.

    Neural Networks

  4. Information Measure Based Skeletonisation
    Sowmya Ramachandran and Lorien Y. Pratt
    Advances in Neural Information Processing Systems Vol. 4., pp. 1080-1087, Denver, Colorado 1992
    Automatic determination of proper neural network topology by trimming over-sized networks is an important area of study, which has previously been addressed using a variety of techniues. In this paper, we present Information Based Skeletonisation (IMBS), a new approach to this problem where superfluous hidden units are removed based on their information measure (IM). This measure, borrowed from decision ttree induction techniques, refelcts the degree to which the hyperplane formed by a hidden unit discriminates between training data classes. We show the results of applying IMBS to three classification tasks and demonstrate that it removes a substantial number of hidden units without significantly affecting network performance.