Fifth International Conference on Artificial Neural Networks (Conf. Publ. No. 440)

7-9 Jul 1997

Filter Results

Displaying Results 1 - 25 of 59
  • A neural relaxation technique for chemical graph matching

    Publication Year: 1997, Page(s):187 - 192
    Cited by:  Papers (2)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (532 KB)

    We develop a binary relaxation scheme for graph matching in chemical databases. The technique works by iteratively pruning the list of matching possibilities for individual atoms based upon contextual information. Its key features include delayed decision-making, robustness to noise, and fast and efficient neural implementation. We illustrate the utility of the technique by comparing it with proba... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural networks for distributed overload control in telecommunications networks

    Publication Year: 1997, Page(s):312 - 317
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (472 KB)

    Overload control in telecom networks is used to protect the network of call processing computers from excessive load during traffic peaks, and involves techniques of predictive control with limited local information. We propose a neural network algorithm, in which a group of neural controllers are trained using examples generated by a globally optimal control method. Simulations show that the neur... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A neural-network approach for compact multi-level representation of human faces

    Publication Year: 1997, Page(s):42 - 47
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (568 KB)

    The authors aim to represent a facial image in a hierarchical manner, extract the most remarkable features on each scale, and select only those which remain most stable for different appearances. This may be used to solve a problem indicated by Mallat (1996) concerning the integration of high-level information into a low-level representation. The subordinate advantage of this recognition scheme is... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Predictive performance for the detection of underfitting in density estimation

    Publication Year: 1997, Page(s):24 - 29
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (416 KB)

    A methodology for model selection and validation is presented. The technique is applied to probability density estimation using Radial Basis Function (RBF) Neural Network. A model is assumed to have a good-fit, if it is able to predict the data. We present a procedure to test if the prediction of the model is calibrated, i.e. if the predicted data frequencies match the empirical data frequencies. ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • From neuronal stochasticity to intelligent resource management of packet data networks

    Publication Year: 1997, Page(s):180 - 186
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (612 KB)

    We explore the neuronal properties of information processing in the real neuron, and extend these toward the optimal and/or intelligent control of resource management within packet type communication networks View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Towards the autonomous control of mobile robots by connectionist experts

    Publication Year: 1997, Page(s):134 - 139
    Cited by:  Patents (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (460 KB)

    This paper concerns an investigation into the form of mobile robot control known as reactive navigation. It describes the development of control architectures that are layered, as in the well known subsumption architecture, but constructed on connectionist principles. Thus individual navigational competences are not hand-crafted but derive from the learning capabilities of the nets employed. In th... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Shadow targets: a novel algorithm for topographic projections by radial basis functions

    Publication Year: 1997, Page(s):7 - 12
    Cited by:  Papers (3)  |  Patents (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (480 KB)

    The archetypal artificial neural network topographic paradigm, Kohonen's self-organising map, has proven highly effective in many applications but nevertheless has significant disadvantages which can limit its utility. Alternative feedforward neural network approaches, including a model called `NEUROSCALE', have been developed based on explicit distance preservation criteria. Excellent generalisat... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A multi-sensor integration method of signals in a metal cutting operation via application of multilayer perceptron neural networks

    Publication Year: 1997, Page(s):306 - 311
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (520 KB)

    The potential application of neural networks in manufacturing scenarios is increasingly becoming feasible. Typical of such manufacturing scenario is the integration of metal cutting sensor signals in pursuance of reliable tool condition monitoring (TCM) system. Successful application of this method of sensor integration could save downtime and costs, that would otherwise not have been realised thr... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • System identification using selforganizing feature maps

    Publication Year: 1997, Page(s):100 - 105
    Cited by:  Papers (2)  |  Patents (2)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (492 KB)

    A method for identification of mechanical systems is reported. The identification of mechanical systems is often done by neural networks used as black boxes in order to produce an inverse system model for control. Contrary to this approach, we intend to identify the mechanical structure and parameters, which allows the use of conventional control theory. The basis of the identification system is a... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Estimations of error bounds for RBF networks

    Publication Year: 1997, Page(s):227 - 232
    Cited by:  Papers (3)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (568 KB)

    The training and optimisation of neural networks to perform function approximation tasks is well documented in the literature. The usefulness of neural networks will be enhanced if a further capacity is added to them: the ability to estimate the accuracy of the results which they generate. Not only will this provide users of neural networks with a confidence index, it will also enable the estimate... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural networks and seasonal time-series prediction

    Publication Year: 1997, Page(s):36 - 41
    Cited by:  Papers (2)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (448 KB)

    Seasonal time series with dominant frequency components are very common in various information and control systems, biomedicine and environmental engineering. Methods for their analysis, modelling and prediction include both linear and nonlinear structures using autoregressive models, adaptive systems and neural networks. The paper presents basic algorithms for preprocessing of such signals includ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Characterising complexity in a radial basis function network

    Publication Year: 1997, Page(s):19 - 23
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (388 KB)

    Attempting to match the complexity of a neural network to the complexity of a data set is difficult as there is no method to determine the effective total degrees of freedom of a network. In this paper we introduce a method for characterising the degrees of freedom of a Radial Basis Function network by exploiting a relationship to the theory of linear smoothers. Specifically, complexity of the mod... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • In-situ ellipsometry solutions using a radial basis function network

    Publication Year: 1997, Page(s):173 - 179
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (560 KB)

    The drive to improve semiconductor device performance has led to a need for advanced semiconductor materials and complex multilayer structures. Fabrication of such devices requires precise control of semiconductor layer parameters such as thickness, composition and interface sharpness. Calculation of thickness and composition from nondestructive ellipsometric measurements is an ill-posed problem w... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • C-NNAP: a dedicated platform for binary neural networks

    Publication Year: 1997, Page(s):161 - 166
    Cited by:  Patents (2)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (488 KB)

    This paper describes techniques for the hardware implementation of a Correlation Matrix Memory (CMM), which is a fundamental element of a binary neural network. For large scale problems the CMM algorithm requires dedicated accelerating hardware to maintain the processing rates required. This paper describes the C-NNAP architecture, which provides processing rates nearly eight times faster than a m... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Predictive terrain contour mapping for a legged robot

    Publication Year: 1997, Page(s):129 - 133
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (372 KB)

    Most legged robots have to negotiate unknown environments with little or no descriptive terrain data as autonomous terrain mapping facilities for legged robots are limited. A predictive terrain contour mapping strategy is proposed which considers the use of feedforward neural networks to predict terrain contours in unstructured environments based on sample data extracted from the walking surface d... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Non-linear speech transition visualization

    Publication Year: 1997, Page(s):257 - 261
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (364 KB)

    Modelling context effects and segmental transitions in speech recognition systems is very important. Explicitly modelling segmental transitions in a RNN framework would circumvent these problems. We present an interesting application of Principal Curves, an algorithm to extract a non-linear summary of p-dimensional data firstly published by Hastie and Stuetzle (1989). The algorithm can be used to ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A comparison of fast training algorithms over two real problems

    Publication Year: 1997, Page(s):1 - 6
    Cited by:  Papers (5)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (484 KB)

    This paper compares the speed of convergence of several popular training algorithms over two real problems. The algorithms compared are backpropagation, SuperSAB, Quickprop, conjugate gradient, RPROP and SASS. This work builds on previous studies, by making use of a benchmark collection Proben1, which is designed to improve the quality of training algorithm comparisons. Statistical significance te... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Using neural networks as part of a system to recognise formations of aircraft

    Publication Year: 1997, Page(s):152 - 157
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (504 KB)

    This paper describes a technique for recognising formations of aircraft from data that has been gathered by a number of independent sensors, then fused together to form a single representation of the environment. The task of recognising formations is formulated as a 3-D deformable template matching problem. The amount and type of deformation allowable by each template is learned from noisy example... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Choosing an appropriate model for novelty detection

    Publication Year: 1997, Page(s):117 - 122
    Cited by:  Papers (11)  |  Patents (7)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (528 KB)

    For novelty detection, a description of normality is learnt by fitting a model to the set of normal examples, and previously unseen patterns are then tested by comparing their novelty score (as defined by the model) against some threshold. Models need to be assessed not only according to their ability to separate normal and abnormal examples but also according to the extent of overfitting relative... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A two-ANN approach to frequency and harmonic evaluation

    Publication Year: 1997, Page(s):245 - 250
    Cited by:  Papers (8)  |  Patents (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (416 KB)

    In power electronics, there are often occasions where the frequency of the signal monitored varies significantly during the measurement period. This creates a huge problem for the conventional spectrum analysis using the fast Fourier transform. An obvious example is the measurement of the output of a variable voltage and variable frequency converter during the motor acceleration/deceleration stage... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Fifth International Conference on Artificial Neural Networks (Conf. Publ. No.440)

    Publication Year: 1997
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (372 KB)

    The following topics were covered: algorithms; signal and image processing; Bayesian networks; self-organising systems; system identification and control; hardware; and applications View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • RAM-based Sigma-pi nets for high accuracy function mapping

    Publication Year: 1997, Page(s):300 - 305
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (380 KB)

    We investigate the use of digital “Higher Order” Sigma-pi nodes and study continuous input RAM-based Sigma-pi units trained with the backpropagation training regime to learn functions to a high accuracy, using these hardware realisable units which may be implemented in microelectronic technology. One of our goals was to achieve accuracies of better than one percent for target output fu... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Magnification factors for the GTM algorithm

    Publication Year: 1997, Page(s):64 - 69
    Cited by:  Papers (7)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (552 KB)

    The generative topographic mapping (GTM) algorithm of C.M. Bishop et al. (1996) has been introduced as a principled alternative to the self-organizing map (SOM). As well as avoiding a number of deficiencies in the SOM, the GTM algorithm has the key property that the smoothness properties of the model are decoupled from the reference vectors, and are described by a continuous mapping from a lower-d... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Self-organizing construction of hierarchical structure of multi-layer perceptrons

    Publication Year: 1997, Page(s):285 - 290
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (420 KB)

    A novel algorithm for creation of a hierarchical structure of neural network classifiers for classification of large databases is suggested. Each node of the hierarchical tree is a multilayer perceptron trained by the algorithm combining self-organization with supervised learning. Thus, the problems of clustering and classification for a given node are solved in concord. Also, it allows the a prio... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Design of a LVQ neural network for compressed image indexing

    Publication Year: 1997, Page(s):94 - 99
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (648 KB)

    This paper proposes an LVQ neural network design to retrieve compressed images from visual databases by content based technology. For image compression, a so called distortion equalized fuzzy learning algorithm is proposed to vector quantize all images before they are stored in the database. For image indexing, a weighted counting of codeword scheme is designed to construct histograms to address t... View full abstract»

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