By Topic

Adaptive Filtering, Non-Linear Dynamics and Neural Networks, IEE Colloquium on

Date 22 Nov 1991

Filter Results

Displaying Results 1 - 11 of 11
  • Algorithms and architectures for nonlinear prediction and filtering

    Publication Year: 1991, Page(s):1/1 - 1/2
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (72 KB)

    Summary form only given. The author uses a particular example to consider the relationship between adaptive filtering, nonlinear dynamics and neural networks. The example is an application of recent developments in nonlinear dynamical systems theory to signal processing based on the radial basis function least squares processor View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Non-linear adaptive equalisation using a switched coefficient adaptive filter

    Publication Year: 1991, Page(s):3/1 - 3/4
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (252 KB)

    Previous non-linear approaches to adaptive filtering problems such as adaptive equalisation have relied upon performing very computationally intensive procedures, compared with linear techniques. The piecewise linear approximation to a non-linear problem is considered. The principal advantage to using the piecewise linear approach is that it can be implemented with virtually no additional computat... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Improved generalization in multi-layer perceptrons with the log-likelihood cost function

    Publication Year: 1991, Page(s):8/1 - 8/4
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (208 KB)

    In supervised training of neural networks, synaptic weights are usually updated by an iterative algorithm which searches for the minimum of some cost function. The most common choice of cost function is the sum of squares (SS). An alternative choice of cost function is the log likelihood (LL). An analytical comparison of the SS and LL has suggested that the latter should lead to improved generaliz... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural network training for a non-linear dynamical system

    Publication Year: 1991, Page(s):5/1 - 5/4
    Cited by:  Papers (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (120 KB)

    The dynamics of a robot are highly nonlinear: for precise positioning of a manipulator, control requires the solution of the inverse dynamics problem for a complex nonlinear system in real time. A neural network offers one of the most promising solutions to these problems. Being a nonlinear dynamical system, a neural network can be trained to learn different relations between variables regardless ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A split-band IIR adaptive line enhancer

    Publication Year: 1991, Page(s):1011 - 1016
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (216 KB)

    A new split-band infinite impulse response (IIR) adaptive line enhancer structure is proposed. This structure is particularly attractive because it overcomes a number of the characteristic problems of IIR adaptive filters, such as instability and the additional complexity needed for the generation of the necessary gradient terms to be used in the adaptive algorithm. It is composed of computational... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • IEE Colloquium on `Adaptive Filtering, Non-Linear Dynamics and Neural Networks' (Digest No.176)

    Publication Year: 1991
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (64 KB)

    The following topics were dealt with: nonlinear prediction; adaptive equalisation; linear Bayesian neurons; neural network training; neural classification; multi-layer perceptrons; filters; and an adaptive IIR line enhancer View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Vector subspaces in non-linear prediction

    Publication Year: 1991, Page(s):2/1 - 2/6
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (196 KB)

    Radial basis function and Volterra series predictors are examined with a view to reducing their complexity while maintaining prediction performance. A geometrical interpretation of the problem is presented. This interpretation indicates that while a multiplicity of choices of reduced state predictor exist, some may be better than others in terms of the numerical conditioning of the solution View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The relationship between Kohonen learning and Kalman filters

    Publication Year: 1991, Page(s):7/1 - 7/5
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (188 KB)

    A considerable amount of research has been devoted to learning mechanisms where a set of input and output pairs are applied to a network, and the network adapts so that given the same set of inputs in the future it will produce the corresponding set of outputs. In practically all of these methods the starting point is Hebbian learning where weights on the inputs are strengthened if the output fire... View full abstract»

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

    Publication Year: 1991, Page(s):4/1 - 4/4
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (172 KB)

    The serious use of neural networks in real-time applications is handicapped by uncertain and possibly lengthy convergence times and unknown final performance. To a large extent these difficulties are removed in the Bayesian neural nets, whose final state is that of a likelihood computer and which have fast training. Whether these are neural networks in the strict sense is debatable but they show m... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive artificial neural network algorithms for MCD applications

    Publication Year: 1991, Page(s):9/1 - 9/8
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (320 KB)

    The MCD (multi-carrier demodulator) has gained interest recently for use in both the aviation and satellite onboard processing contexts, where multiplex format conversions are required. Some experiments indicate that the Rician fading channel is well suited to describe the mobile-satellite channel behaviour. A model for implementing a Rician fading channel is presented. The need for an equaliser i... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Structural generalization in neural classification: incorporation of prior probabilities

    Publication Year: 1991, Page(s):6/1 - 6/5
    Cited by:  Papers (1)  |  Patents (1)
    IEEE is not the copyright holder of this material | Click to expandAbstract | PDF file iconPDF (268 KB)

    Supervised learning of classifications under l2 costing of error trains learning classifiers to recall Bayesian a posteriori probabilities of the possible classes, given observed measurements. This result leads to a number of insights concerning the validation of training, access to the likelihood function, creating networks of networks, incorporation of prior probabilities (wh... View full abstract»

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