By Topic

# 1991 Second International Conference on Artificial Neural Networks

## Filter Results

Displaying Results 1 - 25 of 82
• ### Homotopy continuation method for neural networks

Publication Year: 1991, Page(s):19 - 23
Cited by:  Papers (5)
| | PDF (234 KB)

This paper proposes the use of homotopy continuation method for training the neural networks. Homotopy methods have been demonstrated to be superior to gradient descent methods in many applications such as nonlinear optimization problems, mathematical analysis, and more recently, signal processing. The power of these methods lies in their ability to provide globally convergent solutions, and under... View full abstract»

• ### Operational fault tolerance of the ADAM neural network system

Publication Year: 1991, Page(s):285 - 289
Cited by:  Papers (1)
| | PDF (469 KB)

Neural networks offer a powerful parallel distributed computational system which can be trained to solve many problems. They also appear to be inherently fault tolerant. This is unlike a conventional computing system where fault tolerance is achieved by redundancy, thus increasing its overall complexity. The fault tolerance of the advanced distributed associative memory neural network (ADAM) is in... View full abstract»

• ### The application of neural networks to cognitive phonetic modelling

Publication Year: 1991, Page(s):280 - 284
Cited by:  Papers (1)
| | PDF (176 KB)

A neural network is used to generate control parameters for a parallel formant speech synthesizer, corresponding to a sequence of allophonic tokens. Training is to be accomplished using formant data obtained from both natural and synthetic speech. It is intended that theories of cognitive phonetics, currently being developed in the Department of Language and Linguistics at the University of Essex ... View full abstract»

• ### Building new layers on multi-layer perceptrons

Publication Year: 1991, Page(s):276 - 279
| | PDF (280 KB)

It is widely recognised that the more layers a multi-layer perceptron (MLP) has the longer it takes to train. This is due partly to the larger number of parameters to be calculated, but also to the fact that the back-propagated error is scaled by a number less than unity at each layer. The technique proposed in this paper allows a smaller (fewer layers) network to be created and (partially) traine... View full abstract»

• ### A comparison of connectionist and traditional methods applied to phrase classification and grammaticality determination

Publication Year: 1991, Page(s):190 - 194
Cited by:  Patents (1)
| | PDF (324 KB)

Two sentence classification experiments are described. In the first, the neural net HODYNE is shown to be better than a keyword method at classifying banking phrases. In the second, three techniques are compared at determining the grammaticality of strings of syntactic tags from the LOB corpus: a method using the statistics of tag-pairs, HODYNE, and a multi-layer perceptron. HODYNE and the statist... View full abstract»

• ### Oscillatory neural networks and their application to sensory-motor coordination and control in adaptive robots

Publication Year: 1991, Page(s):328 - 332
| | PDF (304 KB)

The authors research programme has concentrated on two areas: (i) investigation of the dynamic behaviour of networks composed of arrays of coupled nonlinear differential equations, each equation modelling the leaky integrator shunting dynamics of membrane potential; (ii) the development of an initial outline scheme for a sensory-motor coordination and control system for an intelligent robot based ... View full abstract»

• ### A back propagation network as a decision aid in flexible welding system design

Publication Year: 1991, Page(s):271 - 275
| | PDF (304 KB)

A neural network has been applied to a new classification and coding system, a sub-set of Group Technology. The new code matches welding processing requirements of components with the features of a flexible welding cell. This work forms part of the development of a new method of flexible manufacturing systems design. The emphasis of the paper is coding the inputs and outputs of the network, and th... View full abstract»

• ### Self-supervised training of hierarchical vector quantisers

Publication Year: 1991, Page(s):5 - 9
Cited by:  Papers (1)
| | PDF (316 KB)

The author has previously developed a hierarchical vector quantisation (VQ) model which successfully applied to time series and image compression respectively. The paper derives an extension to this model, in which the author backpropagates signals from higher to lower layers of the hierarchy to self-supervise the training of the VQ. He reviews the basic properties of his VQ model and its relation... View full abstract»

• ### Algebraic learning in syntactic neural networks

Publication Year: 1991, Page(s):185 - 189
Cited by:  Patents (2)
| | PDF (368 KB)

The paper presents a new class of learning algorithms for syntactic neural networks (SNN), where each connection weight in the net is represented by a variable, rather than a number. As each training pattern is processed, this leads to an algebraic expression for the output; this expression is then equated to the desired net output. In this manner, the author derives a set of simultaneous equation... View full abstract»

• ### A general purpose neural network architecture for time series prediction

Publication Year: 1991, Page(s):323 - 327
Cited by:  Papers (1)
| | PDF (236 KB)

The paper describes an innovative neural network architecture which is particularly suited to time series prediction applications. The system, which based on a fully connected recurrent network, has been evaluated for both deterministically and stochastically generated time series as well as real process data. Results are presented for the latter and comparisons made against performance achieved b... View full abstract»

• ### Neural processing of airborne sonar for mobile robot applications

Publication Year: 1991, Page(s):267 - 270
Cited by:  Papers (1)
| | PDF (252 KB)

Describes a range of neural signal processing methods employed for B-Scan ultrasonic image enhancement and material identification. All approaches assume no a-priori knowledge of the environment. A Multi-Layered Perceptron (MLP) employing back propagation learning was used for all aspects of this research. The motivation for this work arises from a requirement to map and navigate within, hazardous... View full abstract»

• ### EEG analysis using self-organisation

Publication Year: 1991, Page(s):210 - 213
Cited by:  Papers (1)
| | PDF (216 KB)

The electro-encephalogram (EEG) has formed the basis for the classification of sleep into several stages. The authors propose a method of sleep analysis which requires no pre-defined application of rules, and aims to give some indication of the dynamics of sleep in humans. The authors show that the use of a self-organising feature map has enabled clustering of feature vectors in a high dimensional... View full abstract»

• ### Second International Conference on Artificial Neural Networks (Conf. Publ. No.349)

Publication Year: 1991
| | PDF (24 KB)

The following topics were dealt with: artificial neural networks theory; implementations; images; engineering applications; dynamical systems; control and robotics; hybrids; speech and natural language; medical applications; and character recognition View full abstract»

• ### Robot control using the feedback-error-learning rule with variable feedback gain

Publication Year: 1991, Page(s):139 - 143
Cited by:  Papers (3)
| | PDF (284 KB)

A problem plaguing the application of neural networks in all fields is the difficulty of incorporating prior information to constrain the possible functions that the network can represent. This problem is addressed and solved in a robot control application. The authors present the numerical results of extensive simulations where the main interest is to investigate the generalization ability of the... View full abstract»

• ### A high order feedback net (HOFNET) with variable non-linearity

Publication Year: 1991, Page(s):59 - 63
| | PDF (196 KB)

Most neural networks proposed for pattern recognition sample the incoming image at one instant and then analyse it. This means that the data to be analysed is limited to that containing the noise present at one instant. Time independent noise is therefore, captured but only one sample of time dependent noise is included in the analysis. If however, the incoming image is sampled at several instants... View full abstract»

• ### On the geometry of feedforward neural network weight spaces

Publication Year: 1991, Page(s):1 - 4
Cited by:  Papers (3)
| | PDF (220 KB)

As is well known, many feedforward neural network architectures have the property that their overall input-output function is unchanged by certain weight permutations and sign flips. The existence of these properties implies that if a global optimum of the network performance surface exists at some finite weight vector position, then many copies of a global minimum can be generated by geometric we... View full abstract»

• ### Robot path planning using resistive grids

Publication Year: 1991, Page(s):149 - 152
| | PDF (224 KB)

The use of resistive grids for parallel analogue computation was first suggested by Horn, and has more recently been exploited successfully by Carver Mead and his co-workers at Caltech, for example for building silicon models of the retina. These resistive networks cannot strictly be described as neural networks' in the conventional sense, but they do perform local, parallel, analogue computation... View full abstract»

• ### Logical neural nets and distributed implementations of weighted regular languages

Publication Year: 1991, Page(s):158 - 162
| | PDF (296 KB)

A logical neural network, Aleksander (1), is a finite state machine then it is only possible to recognise regular grammars with these networks. When extra memory is associated with the nodes of these networks, the computational power of the model is increased and now weighted regular grammars, Salomaa (14), can be recognised. Through a constructive method based on the complexity of the production ... View full abstract»

• ### A rule-based dynamic back-propagation (DBP) network

Publication Year: 1991, Page(s):170 - 174
| | PDF (292 KB)

The paper presents and explains experiments performed on a neural network paradigm which works on the Back-Propagation (BP) formula together with additional rules for modifying the net structure. The function of the rules is to control the number of hidden units and their interconnections of a BP net. Hence, the net is capable of evolving' into the optimal topology itself without interference fro... View full abstract»

• ### Fast algorithms to find invariant features for a word recognizing neural net

Publication Year: 1991, Page(s):180 - 184
Cited by:  Papers (1)
| | PDF (284 KB)

A short description of the feature finding neural net (FFNN) for the recognition of isolated words will be given. As has been shown elsewhere, during recognition mode FFNN is faster than the classical HMM and DTW recognizers and yields similar recognition rates. In this article the emphasis is placed on optimal and fast algorithms for selecting relevant features from the speech signal. By the grow... View full abstract»

• ### Improving three layer neural net convergence

Publication Year: 1991, Page(s):318 - 322
| | PDF (292 KB)

The authors investigate the relationship between three layer feed forward back-propagation nets (using the terminology of Rumelhart et al., see Nature vol.323, p.533 et seq., 1986) and the committee net of (Nilsson, see Learning Machines, McGraw-Hill, 1956), and show that a simple modification to the algorithm of the latter makes them, in respect of their power to classify data sets, equivalent. T... View full abstract»

• ### NEUROGEN, musical composition using genetic algorithms and cooperating neural networks

Publication Year: 1991, Page(s):309 - 313
Cited by:  Papers (1)  |  Patents (1)
| | PDF (304 KB)

NEUROGEN has been designed with a view to producing small diatonic, western-type, four-part harmony compositions using the knowledge extracted from a set of example musical fragments provided by the user. The aim has been to produce a piece of coherent music that resembles that typically found in traditional hymns. A set of neural networks are used to capture the conceptual ideas that build good'... View full abstract»

• ### Experience in using neural networks for electronic diagnosis

Publication Year: 1991, Page(s):115 - 118
Cited by:  Papers (5)  |  Patents (3)
| | PDF (244 KB)

British Telecommunication plc (BT) has an interest in developing fast, efficient diagnostic systems especially for high volume circuit boards as found in today's digital telephone exchanges. Previous work to produce a diagnostic system for line cards has shown that a model-based, expert system shell can be most beneficial in assisting in the diagnosis and subsequent repair of these complex, mixed-... View full abstract»

• ### Universal architectures for logical neural nets

Publication Year: 1991, Page(s):262 - 266
Cited by:  Papers (1)
| | PDF (284 KB)

A universal architecture of logical neural nets is proposed which includes the conventional N-tuple and pyramid architectures as its extremes. The discrimination function of the architecture can be adjusted conveniently via the structure parameters and proper spreading operation. This flexibility enables tailored discriminator design in a practical environment. The technique of spreading with a mu... View full abstract»

• ### Estimation and neurocontrol in the presence of feedback

Publication Year: 1991, Page(s):300 - 303
| | PDF (228 KB)

When feedforward adaptive control is applied to distributed parameter plants a feedback path often exists around the controller. This feedback path introduces a potential instability and greatly changes the behaviour of the controlled system, usually degrading performance with respect to the no feedback' case. The feedback induced effects present in feedforward neurocontrol are similar to those e... View full abstract»