1999 Ninth International Conference on Artificial Neural Networks ICANN 99. (Conf. Publ. No. 470)

1999

Go

Filter Results

Displaying Results 1 - 25 of 90
  • Time-slicing: a model for cerebellar function based on synchronization, reverberation, and time windows

    Publication Year: 1999, Page(s):539 - 544 vol.2
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (484 KB)

    We present a new theory of cerebellar function that is based on synchronization, delayed reverberation, and time windows for triggering spikes. We show that granule cells admit messy fiber activity to the parallel fibers only if the Golgi cells are firing synchronously and if the spikes arrive within short and well-defined time windows. The time window mechanism organizes neuronal activity in disc... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Backtracking deterministic annealing for constraint satisfaction problems

    Publication Year: 1999, Page(s):868 - 873 vol.2
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (488 KB)

    We present a deterministic annealing approach to the solution of quadratic constraint satisfaction problems with complex interlocking constraints, such as exemplified in polyomino tiling puzzles. We first analyze the dynamical properties of the solution strategies implemented by deterministic annealing (DA) in the analog neural representation of Potts-mean-field (PMF) and penalty-function-based co... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Regularisation of RBF-networks with the Bayesian evidence scheme

    Publication Year: 1999, Page(s):533 - 538 vol.2
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (364 KB)

    We propose a novel regularisation method for Gaussian multiple networks, which adopts a Bayesian approach and draws on the evidence scheme to optimise the hyperparameters. This leads to a new, modified form of the EM algorithm, which is compared with the original scheme on three classification problems View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural network training using multi-channel data with aggregate labelling

    Publication Year: 1999, Page(s):862 - 867 vol.2
    Cited by:  Papers (2)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (420 KB)

    The solution of classification problems using statistical techniques requires appropriately labelled training data. In the case of multi-channel data, however, the labels may only be available in aggregate form rather than as separate labels for each individual-channel. Standard techniques, using a trained model to classify each channel separately, are therefore precluded. We present a method of t... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • SDTs: sparse dynamic trees

    Publication Year: 1999, Page(s):527 - 532 vol.2
    Cited by:  Papers (4)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (456 KB)

    We introduce a class of image models which we call sparse dynamic trees (SDTs). These extend our previous work (1999) on dynamic trees by introducing a top-down generative prior for the tree structures, which leads to a sparse use of nodes in the network. We present results showing the properties of these networks in action on 1D patterns View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • How irrelevant inputs affect MLP pattern based learning

    Publication Year: 1999, Page(s):856 - 861 vol.2
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (368 KB)

    We analyze the effect of irrelevant inputs on pattern based learning dynamics of multilayer perceptrons (MLPs). We propose a model for the evolution of weights linking these irrelevant inputs to hidden network nodes. The model is analyzed for a restricted case and the results are illustrated and discussed View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Simultaneous Lp-approximations of polynomials and derivatives on the whole space

    Publication Year: 1999, Page(s):587 - 592 vol.2
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (328 KB)

    We have obtained a sufficient condition that a linear sum of an activation function can simultaneously approximate polynomials and their derivatives in the sense of Lp(Rd, μ). If the probability measure μ is rapidly decreasing, a wide variety of differentiable functions satisfy this condition. For the Gaussian measure, rapidly increasing functions such as et,... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Generative vector quantisation

    Publication Year: 1999, Page(s):934 - 939 vol.2
    Cited by:  Papers (2)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (420 KB)

    Based on the assumption that a pattern is constructed out of features which are either fully present or absent, we propose a vector quantisation method which constructs patterns using binary combinations of features. For this model there exists an efficient EM-like learning algorithm which learns a set of representative codebook vectors. In terms of a generative model, the collection of allowed bi... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Selecting features in neurofuzzy modelling by multiobjective genetic algorithms

    Publication Year: 1999, Page(s):749 - 754 vol.2
    Cited by:  Papers (3)  |  Patents (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (476 KB)

    Empirical modelling in high dimensional spaces is usually preceded by a feature selection stage. Irrelevant or noisy features unnecessarily increase the complexity of the problem and can degrade modelling performance. Here, multiobjective genetic algorithms are proposed as effective means of evolving a diverse population of alternative feature sets with various accuracy/complexity trade-offs. They... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Regularisation of mixture density networks

    Publication Year: 1999, Page(s):521 - 526 vol.2
    Cited by:  Papers (2)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (412 KB)

    Mixture density networks (MDNs) are a well-established method for modelling complex multi-valued functions where regression methods (such as MLPs) fail. In this paper we develop a Bayesian regularisation method for MDNs by an extension of the evidence procedure. The method is tested on two data sets and compared with early stopping View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Learning to forget: continual prediction with LSTM

    Publication Year: 1999, Page(s):850 - 855 vol.2
    Cited by:  Papers (5)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (448 KB)

    Long short-term memory (LSTM) can solve many tasks not solvable by previous learning algorithms for recurrent neural networks (RNNs). We identify a weakness of LSTM networks processing continual input streams without explicitly marked sequence ends. Without resets, the internal state values may grow indefinitely and eventually cause the network to break down. Our remedy is an adaptive “forge... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The effects of intrinsic noise on pattern recognition in a model pyramidal cell

    Publication Year: 1999, Page(s):1006 - 1011 vol.2
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (412 KB)

    Computer simulation of a CA1 hippocampal pyramidal cell is used to estimate the effects of synaptic and spatio-temporal noise on signal integration. Comparison is made between the pattern recognition ability of the cell in the presence of this noise and that of a computing unit in an artificial neural network model of an hetero-associative memory. The results indicate that the pattern recognition ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Beyond independent components

    Publication Year: 1999, Page(s):809 - 814 vol.2
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (760 KB)

    Independent component analysis (ICA) attempts to find a linear decomposition of observed data vectors into components that are statistically independent. It is well known, however, that such a decomposition cannot be exactly found, and in many practical applications, independence is not achieved even approximately. This raises the question on the utility and interpretation of the components given ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Maximizing information about a noisy signal with a single non-linear neuron

    Publication Year: 1999, Page(s):581 - 586 vol.2
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (484 KB)

    For noise-free information maximization, the output signal entropy must be maximized. This is not true for a noisy input: rather, it must be the difference between this entropy and the residual output uncertainty. A definition of information density is introduced, which provides a discrete local measure of bandwidth efficiency. Novel training rules are proposed which enforce a uniformity of this d... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An analysis of initial state dependence in generalized LVQ

    Publication Year: 1999, Page(s):928 - 933 vol.2
    Cited by:  Papers (5)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (348 KB)

    The author proposed a new formulation of learning vector quantisation (LVQ) called generalized LVQ (GLVQ) based on the minimum classification error criterion. In this paper, the initial state dependence in GLVQ is discussed, and it is clarified that the learning rule should be modified to make it insensitive to the initial values of reference vectors. The robustness of the modified GLVQ for the in... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Learning error-correcting output codes from data

    Publication Year: 1999, Page(s):743 - 748 vol.2
    Cited by:  Papers (6)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (392 KB)

    A polychotomizer which assigns the input to one of K⩾3 classes is constructed using a set of dichotomizers which assign the input to one of two classes. Defining classes in terms of the dichotomizers is the binary decomposition matrix of size K×L where each of the K⩾3 classes is written as error-correcting output codes (ECOC), i.e., an array of the responses of binary decisions made ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Rule extraction from binary neural networks

    Publication Year: 1999, Page(s):515 - 520 vol.2
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (408 KB)

    A new constructive learning algorithm, called Hamming clustering (HC), for binary neural networks is proposed. It is able to generate a set of rules in if-then form underlying an unknown classification problem starting from a training set of samples. The performance of HC has been evaluated through a variety of artificial and real-world benchmarks. In particular, its application in the diagnosis o... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Minimum entropy data partitioning

    Publication Year: 1999, Page(s):844 - 849 vol.2
    Cited by:  Papers (8)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (380 KB)

    Problems in data analysis often require the unsupervised partitioning of a data set into clusters. Many methods exist for such partitioning but most have the weakness of being model-based (most assuming hyper-ellipsoidal clusters) or computationally infeasible in anything more than a 3D data space. We re-consider the notion of cluster analysis in information-theoretic terms and show that minimisat... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Influence of dendritic morphology on axonal competition

    Publication Year: 1999, Page(s):1000 - 1005 vol.2
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (372 KB)

    The development of nerve connections involves competition among axons for survival promoting factors, or neurotrophins, which are released by the axons' target cells. To study the influence of the target's dendritic tree on axonal competition, the authors have extended their previous model of axonal competition (1999) to take into account the extracellular space around the dendrites. The authors s... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • VC dimension bounds for higher-order neurons

    Publication Year: 1999, Page(s):563 - 568 vol.2
    Cited by:  Papers (4)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (448 KB)

    We investigate the sample complexity for learning using higher-order neurons. We calculate upper and lower bounds on the Vapnik-Chervonenkis dimension and the pseudo dimension for higher-order neurons that allow unrestricted interactions among the input variables. In particular, we show that the degree of interaction is irrelevant for the VC dimension and that the individual degree of the variable... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Piecewise affine neural networks and nonlinear control: stability results

    Publication Year: 1999, Page(s):608 - 612 vol.2
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (228 KB)

    Concerns stability properties of artificial neural networks tuned to nonlinear control. The neural networks used are piecewise affine perceptrons (PAP), a subclass of perceptrons. They have properties that can be used to initialize them to control a given nonlinear system. Besides they have the same useful properties as classical perceptrons: the universal approximation property and the generaliza... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • MLP emulation of N-gram models as a first step to connectionist language modeling

    Publication Year: 1999, Page(s):910 - 915 vol.2
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (536 KB)

    In problems such as automatic speech recognition and machine translation, where the system response must be a sentence in a given language, language models are employed in order to improve system performance. These language models are usually N-gram models (for instance, bigram or trigram models) which are estimated from large text databases using the occurrence frequencies of these N-grams. Nakam... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Response of an excitatory-inhibitory neural network to external stimulation: an application to image segmentation

    Publication Year: 1999, Page(s):803 - 808 vol.2
    Cited by:  Patents (6)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (380 KB)

    Neural network models comprising elements which have exclusively excitatory or inhibitory synapses are capable of a wide range of dynamic behavior, including chaos. In this paper, a simple excitatory-inhibitory neural pair, which forms the building block of larger networks, is subjected to external stimulation. The response shows transition between various types of dynamics, depending upon the mag... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A new adaptive architecture: Analogue synthesiser of orthogonal functions

    Publication Year: 1999, Page(s):714 - 719 vol.2
    Cited by:  Papers (1)  |  Patents (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (312 KB)

    A new adaptive nonlinear (neural-like) architecture, an analogue synthesiser of orthogonal functions which is able to produce a plurality of mutually orthogonal signals as functions of time such as Legendre, Chebyshev and Hermite polynomials, cosine basis of functions, smoothed cosine basis, etc., is proposed. A proof-of-concept breadboard version of the analogue synthesiser is described. The devi... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive margin support vector machines for classification

    Publication Year: 1999, Page(s):880 - 885 vol.2
    Cited by:  Papers (10)  |  Patents (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract |PDF file iconPDF (392 KB)

    We propose a learning algorithm for classification learning based on the support vector machine (SVM) approach. Existing approaches for constructing SVMs are based on minimization of a regularized margin loss where the margin is treated equivalently for each training pattern. We propose a reformulation of the minimization problem such that adaptive margins for each training pattern are utilized, w... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.