[Proceedings] 1991 IEEE International Joint Conference on Neural Networks

18-21 Nov. 1991

Filter Results

Displaying Results 1 - 25 of 444
  • Operational fault tolerance of the ADAM neural network system

    Publication Year: 1991, Page(s):1 - 6 vol.1
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (357 KB)

    The author investigates the fault tolerance of the Advanced Distributed Associative Memory (ADAM), focusing on its operational use. The effect of the reliability of recall of variously configured ADAM systems is examined by injecting faults individually, and also in various combinations since correlations between them will influence their overall effect on the system. Analysis of the results indic... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A general purpose neurocomputer

    Publication Year: 1991, Page(s):361 - 366 vol.1
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (252 KB)

    Presents a neural network, composed of linear units with threshold, as the CPU of a stored program MIMD architecture. The Caianiello formalism, is introduced as an aid to implement the arithmetic and control algorithms, needed for the smooth running of this general-purpose system. That is, in the neural net both the arithmetic and logic algorithms and the operating system have been implemented. Th... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Building up neuromimetic machines with LNeuro 1.0

    Publication Year: 1991, Page(s):602 - 607 vol.1
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (346 KB)

    The state of experiments on neural networks simulations on a parallel architecture is presented. The computing device, LNeuro 1.0, is based on an existing coarse-grain parallel framework (INMOS Transputers), improved with finer grain parallel abilities through VLSI modules. A digital architecture, scalable and flexible enough to be useful for simulating various kinds of networks and paradigms, was... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • 1991 IEEE International Joint Conference on Neural Networks (Cat. No.91CH3065-0)

    Publication Year: 1991
    Request permission for commercial reuse | |PDF file iconPDF (47 KB)
    Freely Available from IEEE
  • Nonlinear neural field filters

    Publication Year: 1991, Page(s):1885 - 1890 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (132 KB)

    Design, stability and implementation of nonlinear neural field filters are examined. The input and output of the neural field filters are vector fields. A neural transform is used to represent the input, output signals and the transfer function of the neural field filter. It is concluded that the Lyapunov conditions for such fields are taken care of by a novel extension of the Routh stability crit... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Approximations of mappings and application to translational invariant networks

    Publication Year: 1991, Page(s):2294 - 2298 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (209 KB)

    The author studies the approximation of continuous mappings and dichotomies by one-hidden-layer networks, from a computational point of view. The approach is based on a new approximation method, specially designed for constructing small networks. Upper bounds are given on the size of these networks. These results are specialized to the case of transitional invariant networks, i.e., networks whose ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Improving the training speed of three-layer feedforward neural nets by optimal estimation of the initial weights

    Publication Year: 1991, Page(s):2063 - 2068 vol.3
    Cited by:  Papers (18)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (244 KB)

    The authors formulate the training problem for three-layer feedforward neural nets based on the well known linear algebra of D. Rumelhart et al. (1986). Then, they develop two estimation algorithms, called the forward estimation algorithm and the recurrent estimation algorithm, to estimate the initial weights. The basic idea is to set the initial weights space as close as possible to a global mini... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Efficient gradient computation for continuous and discrete time-dependent neural networks

    Publication Year: 1991, Page(s):2337 - 2342 vol.3
    Cited by:  Papers (4)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (232 KB)

    The author provides calculus-of-variations techniques for the construction of backpropagation-through-time (BTT) algorithms for arbitrary time-dependent recurrent neural networks with both continuous and discrete dynamics. The backpropagated error signals are essentially Lagrange multipliers. The techniques are easy to handle because they can be embedded into the Hamiltonian formalism widely used ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural training and generalisation of sequences using continuous temporal structure

    Publication Year: 1991, Page(s):2027 - 2032 vol.3
    Cited by:  Papers (2)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (264 KB)

    An approach for sequential neural behavior called continuous backpropagation is evolved from standard backpropagation where a state is replaced by a state transition sequence as the goal weight condition. The approach may be used to train mappings of analog input/output (I/O) signals or discrete I/O sequences with underlying continuity. An arbitrarily increasing number of values in the I/O sequenc... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Signal representation by generalized nonorthogonal Gaussian wavelet groups using lattice networks

    Publication Year: 1991, Page(s):968 - 973 vol.2
    Cited by:  Papers (10)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (300 KB)

    The authors describe a general method for signal representation using nonorthogonal basis functions that are composed of Gaussians. The Gaussians can be combined into groups with predetermined configuration that can approximate any desired basis function. The same configuration at different scales forms a set of self-similar wavelets. The general scheme is demonstrated by representing a natural si... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Learning neural network weights using genetic algorithms-improving performance by search-space reduction

    Publication Year: 1991, Page(s):2331 - 2336 vol.3
    Cited by:  Papers (10)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (236 KB)

    The authors present a technique for reducing the search-space of the genetic algorithm (GA) to improve its performance in searching for the globally optimal set of connection-weights. They use the notion of equivalent solutions in the search space, and include in the reduced search-space only one solution, called the base solution, from each set of equivalent solutions. The iteration of the GA con... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Improvement of robot control by neural computers

    Publication Year: 1991, Page(s):1837 - 1843 vol.2
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (332 KB)

    Currently available robot control systems have various limitations in comparison to biological motor systems partly due to a lack of a general control theory for robots in a dynamic environment and partly due to the real-time challenge for conventional computers. The authors review current approaches to (1) speeding up the neural computation by means of adaptive load distribution on massively para... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Markov random field based image labeling with parameter estimation by error backpropagation

    Publication Year: 1991, Page(s):962 - 967 vol.2
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (308 KB)

    The authors investigate a method of efficiently labeling images using the Markov random field (MRF). The MRF model is defined on the region adjacency graph and the labeling is then optimally determined using simulated annealing. The MRF model parameters are automatically estimated using an error backpropagation network. The proposed method is analyzed through experiments using real natural scene i... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Automatic target recognition using image and network decomposition

    Publication Year: 1991, Page(s):281 - 286 vol.1
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (292 KB)

    A constrained version of the general automatic target recognition problem is formulated in the context of pattern recognition, as the detection and localization of target patterns in scene images. An approach to the simplification of the learning involved is presented. The two-stage network architecture that results from this approach incorporates two cascaded modules: a multilayered perceptron fo... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A layered architecture for regularization vision chips

    Publication Year: 1991, Page(s):1007 - 1020 vol.2
    Cited by:  Papers (2)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (452 KB)

    The authors propose a layered architecture for regularization problems with higher order smoothness constraints which requires only immediate neighborhood wiring and demands no negative conductance. They describe the architecture and show how the network naturally solves regularization problems. They also present an application to the smoothing-contrast enhancement filter for image processing. The... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Autonomous trajectory generation of a biped locomotive robot

    Publication Year: 1991, Page(s):1983 - 1988 vol.3
    Cited by:  Papers (3)  |  Patents (4)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (256 KB)

    Introduces a hierarchical structure for motion planning and learning control of a biped locomotive robot. In this system, trajectories are obtained for a robot's joints on a flat surface by an inverted pendulum equation and a Hopfield type neural network. The former equation is simulated for the motion of the center of gravity of the robot and the network is used for solving the inverse kinematics... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Computer aided investigations of artificial neural systems

    Publication Year: 1991, Page(s):2325 - 2330 vol.3
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (364 KB)

    An attempt is made to demonstrate how symbolic computation can be applied to aid in the analysis and derivation of neural systems. The authors review the general method and techniques of the Lyapunov method for the stability analysis of artificial neural systems. They present some strategies for using computer algebra systems and their extensions to analyze the stability of known neural systems an... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Model reference neurocontrollers based on feedback linearization

    Publication Year: 1991, Page(s):1813 - 1818 vol.2
    Cited by:  Papers (2)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (292 KB)

    The authors report experimental results on learning of the feedback linearizing control laws for an inverted pendulum system based on an unsupervised learning control scheme. Only the case where no state transformation is required for linearizing the nonlinear system is considered. A method inspired by a direct adaptive control scheme was used to learn the linearizing law using artificial neural n... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Inverse modeling of dynamical system-network architecture with identification network and adaptation network

    Publication Year: 1991, Page(s):566 - 571 vol.1
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (272 KB)

    The authors describe a neural network architecture enabling inverse modeling of a nonlinear dynamical system. It consists of two neural networks, a system identification network and an adaptation network. The effectiveness of the proposed network architecture is examined by applying it to a digital mobile communication adaptive equalizer. In digital mobile communication, the problem of multipath f... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A new approach to the design of Hopfield associative memory

    Publication Year: 1991, Page(s):1705 - 1710 vol.2
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (300 KB)

    The authors present a novel method for constructing the weight matrix for the Hopfield associative memory. The most important feature of this method is the explicit introduction of the size of the attraction basin to be a main design parameter, and the weight matrix is obtained as a result of optimizing this parameter. Another feature is that all the connection weights can only assume three differ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Fuzzy feature extraction using a class of neural network

    Publication Year: 1991, Page(s):1618 - 1624 vol.2
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (384 KB)

    The authors present a novel approach to feature extraction using a class of neural networks for the purpose of authorship recognition. The framework of the research is based on the factor space theory proposed by P.Z. Wang (1990). The main advantage of this approach compared to others is that the dimension of the state space required to distinguish the output patterns for a particular recognition ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Adaptive decision-feedback equalization of digital transmission channels using forward-only counterpropagation networks

    Publication Year: 1991, Page(s):220 - 225 vol.1
    Cited by:  Papers (4)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (240 KB)

    An application of the forward-only counterpropagation network (FCPN) is proposed for nonlinear equalization of digital transmission channels. The learning mechanism of the FCPN is a combination of unsupervised self-organization and supervised training. A decision-feedback equalizer (DFE) based on FCPN was simulated on a digital computer. The results of the simulation demonstrate a superior bit-err... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Tabu learning: a neural network search method for solving nonconvex optimization problems

    Publication Year: 1991, Page(s):953 - 961 vol.2
    Cited by:  Papers (3)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (416 KB)

    The authors present a novel technique, called tabu learning, for solving nonconvex optimization problems using neural networks. Tabu learning applies the concept of tabu search to neural networks by continuously increasing the energy surface in a neighborhood of the current state, thus penalizing states already visited. This enables the state trajectory to climb out of local minima while tending t... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A proposal for a hierarchical MRF model based on conditional probability

    Publication Year: 1991, Page(s):268 - 274 vol.1
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (272 KB)

    The standard regularization theory extended to problems where generic constraints or knowledge are expressed within the framework of a Markov random field (MRF) model. This extended theory is applied to image restoration in which a desired state in the line process is given as a constraint. The forward process in transformation between two kinds of visual information, from information of pixel int... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Conditions for robust stability of analog VLSI implementation of neural networks with uncertain circuit parasitics

    Publication Year: 1991, Page(s):1879 - 1884 vol.3
    Request permission for commercial reuse | Click to expandAbstract |PDF file iconPDF (184 KB)

    An analog VLSI implementation of neural networks has been modeled in terms of active cell impedance connected to a resistive grid. The resistive grid can be characterized in terms of the nominal linear component and the parasitic component with uncertain parametric values. Necessary and sufficient conditions for the nominal and robust stability of these systems can then be derived View full abstract»

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