By Topic

Proceedings of 1993 International Conference on Neural Networks (IJCNN-93-Nagoya, Japan)

25-29 Oct. 1993

Go

Filter Results

Displaying Results 1 - 25 of 242
  • Self-organizing neural network for multidimensional mapping and classification of multiple valued data

    Publication Year: 1993, Page(s):2488 - 2491 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (180 KB)

    A new self-organizing neural network for the recognition and prediction of multiple-valued patterns is introduced. It is a supervised learning system which incorporates two MVL ART modules (ARTa and ARTb) that can learn to predict a prescribed m-dimensional output vector given a prescribed n-dimensional input vector. These MVL ART modules are linked via an inter ART mapfield ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Extended backpropagation for invariant pattern recognition neural networks

    Publication Year: 1993, Page(s):2097 - 2100 vol.3
    Cited by:  Patents (9)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (336 KB)

    This paper presents some improvements on a neural network structure composed by a multilayer perceptron (MLP) with a preprocessing neural net, in order to perform translation, rotation and scale invariant pattern recognition. The preprocessing network has been modified and backpropagation (BP) has been generalized for training the preprocessing net as well as the multilayer perceptron. The new str... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Biologically-inspired artificial neurons: modeling and applications

    Publication Year: 1993, Page(s):2300 - 2303 vol.3
    Cited by:  Papers (5)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (320 KB)

    Currently used neural networks employ mostly simple neuron models that greatly differ from the "real" biological neurons. To ensure progress in biology-based neural processing, more advanced neuron models must be developed that better reflect the biological functionality. In this paper, we investigate a neuron model which satisfies such requirements to a much higher degree. We also examine some of... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural learning for adaptive internal model control

    Publication Year: 1993, Page(s):2771 - 2774 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (404 KB)

    This work describes how artificial neural networks can be applied in an adaptive control context. An attempt is made to merge conventional adaptive control concepts with today's neural way of thinking. Thus it is possible to throw light on some obvious links often remaining unnoticed. Special emphasis is put on the learning behavior of the network. Two learning rules are analyzed and tested in a s... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Recurrent neural networks and Fibonacci numeration system

    Publication Year: 1993, Page(s):2331 - 2334 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (216 KB)

    It is known from Zeckendorf's theorem (1972) that every positive integer admits a representation as a sum of distinct Fibonacci numbers. Furthermore, this representation is unique if it does not contain two consecutive digits that equal to 1 and has no zero to its left hand side. This unique representation is called normal form. Recurrent neural networks have shown to have powerful capabilities fo... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Neural algorithms for placement problems

    Publication Year: 1993, Page(s):2421 - 2424 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (328 KB)

    Two improved neural algorithms are presented for solving a placement problem which is a familiar class of NP-hard quadratic assignment problems. Formulation of the problem as a zero-one integer programming leads to an improved form of the Hopfield networks, while a mixed integer programming formulation results in an analogue algorithm similar to the elastic nets. The outermost loop in these algori... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Construction of a feature set for character recognition

    Publication Year: 1993, Page(s):2153 - 2156 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (272 KB)

    A character recognition system which utilized neural network was trained to recognize handwritten alphanumeric characters. Unsupervised learning was adopted in order to ensure the necessary features for recognition were determined by the network in accordance with its structure. An algorithm which forced the network to pick up features with simple shapes was introduced so that the feature extracti... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A self-organizing supervised classifier

    Publication Year: 1993, Page(s):2484 - 2487 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (212 KB)

    A new supervised neural network classifier for online learning is introduced. An association of prototype neurons and fuzzy membership function (MF) is used for cluster approximation. The new architecture based on adaptive resonance theory (ART) dedicates one adapted ART module (ARTMOD) to each class of patterns. Each prototype neuron defines a hyper-sphere in the input space. A class consists of ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Temporal association realized by a network of bursting neurons

    Publication Year: 1993, Page(s):2359 - 2362 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (268 KB)

    Using a neuron model which differs from traditional ones in that individual neurons possess internal structure, we can construct a network having higher dynamical flexibility. From this point of view, we propose a network which has the ability to retrieve not only a sequence of patterns but also their rhythms. In other words, the duration for which each pattern is presented and the duration of eac... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • An identification of human faces using bright-spots matrix projection

    Publication Year: 1993, Page(s):2093 - 2096 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (292 KB)

    In this paper, the authors describe a method for the identification of human faces. In this method, the fiber grating (FG) vision sensor which has been developed by the authors is employed for the three dimensional shape of the faces. Before the identification of the face using the three dimensional shape of the face, it is necessary to calibrate the position and direction of the facial data. In t... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Thermal image processing using neural network

    Publication Year: 1993, Page(s):2065 - 2068 vol.3
    Cited by:  Papers (1)  |  Patents (3)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (264 KB)

    This paper proposes a thermal image processing system that consists of a small, lightweight noncooled thermal image sensor, a thermal image pre-processor, and a structured neural network. This system can measure precisely the position and posture of occupants from feature data using a low-resolution thermal image captured in the polar coordinate system. This thermal image processing system is expe... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A comparison of separating capabilities of the feed-next and the feed-forward three layer neural networks

    Publication Year: 1993, Page(s):2296 - 2299 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (164 KB)

    To implement every dichotomy on a 2M-point set in Rn, the feed-next three layer neural net needs 2M-1 signum neurons in its hidden layer. It is shown that the feedforward neural net needs only M-1 hidden neurons to do it. An active use of an equivalent transformation theorem on the feedforward net to the feed-next net is discussed. View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The comparison research of robot control using BP and CMAC neural network

    Publication Year: 1993, Page(s):2767 - 2770 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (396 KB)

    Discusses two manipulator control strategies using neural nets, BP and CMAC. No prior knowledge of robot dynamics and environments is required when designing the NN controller, because the BP and CMAC have the ability of self-organization and self-learning. The controller consists of two parts,one is a conventional feedback loop, it forms the primary control torque, another is a NN compensator, it... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A global stable analysis for CGNN and CNN with asymmetric weights

    Publication Year: 1993, Page(s):2327 - 2330 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (156 KB)

    We consider the CGNN model of neural networks (Cohen and Grossberg, 1983) and cellular neural network (CNN) model (Yang and Chua, 1988) with asymmetric weights. Using Lasalle's invariance principle, we proved that if the weight matrix in CGNN can be decomposed as the product of a symmetric matrix and a positively definite diagonal matrix, then all bounded orbits of the above model converge to equi... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A parallel implementation of Boltzmann machine solving combinatorial optimization problems

    Publication Year: 1993, Page(s):3046 - 3049 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (356 KB)

    In this paper we present an efficient way to implement the Boltzmann machine for solving combinatorial optimization problems on a distributed-memory multiprocessor (DMM), especially on a network of transputers. In this scheme, the neurons in a Boltzmann machine are partitioned into p disjoint sets and mapped onto each processor, where p is the number of processors in a DMM. Some experimental speed... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • On an optimal learning scheme for bidirectional associative memories

    Publication Year: 1993, Page(s):2670 - 2673 vol.3
    Cited by:  Papers (2)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (268 KB)

    An optimal learning scheme is proposed for a class of bidirectional associative memories (BAMs). This scheme, based on the perceptron learning algorithm, is motivated by the inadequacies/incompleteness of the weighted learning by global optimization, as derived by Wang et al. (1993). It is shown that the new scheme has superior properties: (1) Convergence to the correct solution, when it exists; a... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Run-time robot planning

    Publication Year: 1993, Page(s):2815 - 2818 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (232 KB)

    We (1993) have developed a neural network architecture which learns a forward model of a redundant manipulator (via self-supervised training) as a map of normalized radial basis neurons and inverts the model by means of run-time gradient descent of a task-related potential field. In this paper, we propose a distributed model for the computation of the field, which is consistent with the model-inve... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • The design of multiresolution adaptive classifier

    Publication Year: 1993, Page(s):2181 - 2184 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (232 KB)

    The characteristics of visual information processing at the primary stage of the visual nerve system (from retina to striate cortex) are summarized briefly in this paper. The concept of wavelet transform (WT) implied in this procedure is emphasized. A multiresolution adaptive classifier (MAC)is presented to organize this concept into the network structure and the back propagation (BP) learning met... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Dynamical process of learning chaotic time series by neural networks

    Publication Year: 1993, Page(s):2387 - 2390 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (192 KB)

    We report the result of computer simulations on the learning process of temporal series by artificial neural networks. In our simulation, we used a feedforward neural network model with 4-layers to study the capability and dynamical learning process of chaotic time series produced by triangular maps. We found a critical time (tcr) at which the learning process proceeds abruptly. We also... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • 10000 cell placement optimization using a self-organizing map

    Publication Year: 1993, Page(s):2417 - 2420 vol.3
    Cited by:  Papers (2)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (328 KB)

    A new approach for the cell placement problem using a self-organizing map is proposed. This method requires a memory size of only O(N) to solve an N cell problem. Large scale problems can be solved on a workstation in a reasonable computation time. Simulation results show the method described gives better performance than two conventional methods: the neural method using a feedback type neural net... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Aimpoint selection-a heterogeneous neural network approach

    Publication Year: 1993, Page(s):2149 - 2152 vol.3
    Cited by:  Papers (1)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (352 KB)

    Computer vision is playing an ever more critical role in the expanding world of computer automation. Neural network algorithms have promised to increase the performance and amount of processing that can be done by computer vision systems by reducing the complexity of image processing algorithms and by reducing the amount of time required to produce these image processing algorithms. While homogene... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • A weighted competitive learning extracting skeleton structure from character patterns with non-uniform width

    Publication Year: 1993, Page(s):2480 - 2483 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (332 KB)

    In the handwritten character recognition, it is very important to extract essential structure of character patterns. Requirements for skeletonization can be summarized as follows: (a) Insensitive to irregular edge lines. (b) Nonstructure patterns are not extracted. (c) Insensitive to nonuniform line width. (d) Line information is held. In this paper, a weighted competitive learning method is propo... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Order and rank of neural networks

    Publication Year: 1993, Page(s):2355 - 2358 vol.3
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (304 KB)

    Any boolean function can be expressed as a higher order threshold function. This means that logic networks may be treated as higher order neural networks. The author defines two complexity indices of logic networks when they are viewed as higher order neural networks. One is an order index: this reflects degree of parallelism employed at each element in processing information. The other is a rank ... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • New regular moment invariants to classify elongated and contracted images

    Publication Year: 1993, Page(s):2089 - 2092 vol.3
    Cited by:  Papers (3)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (176 KB)

    This paper presents a technique to classify images that have been elongated or contracted. It is first shown that the conventional regular moment invariant remains no longer invariant when the image is scaled unequally in the x-and y-directions. A method is proposed to form moment invariants that do not change under such unequal scaling. Results of computer simulations for images are also included... View full abstract»

    Full text access may be available. Click article title to sign in or learn about subscription options.
  • Multiply descent cost competitive learning as an aid for multimedia image processing

    Publication Year: 1993, Page(s):2061 - 2064 vol.3
    Cited by:  Papers (3)  |  Patents (4)
    Request permission for commercial reuse | Click to expandAbstract | PDF file iconPDF (416 KB)

    An integration of neural and ordinary computations toward multimedia processing is presented. The handled media is a combination of still images and animations. The neurocomputation here is the multiply descent cost competitive learning. This algorithm generates two types of feature maps. One of them: an optimized grouping pattern of pixels by self-organization, is used. A data-compressed still im... View full abstract»

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