Neural Network Structure (neural + network_structure)

Distribution by Scientific Domains

Selected Abstracts

Natural gradient algorithm for neural networks applied to non-linear high power amplifiers,

H. Abdulkader
Abstract This paper investigates the processing techniques for non-linear high power amplifiers (HPA) using neural networks (NNs). Several applications are presented: Identification and Predistortion of the HPA. Various Neural Network structures are proposed to identify and predistort the HPA. Since a few decades, NNs have shown excellent performance in solving complex problems (like classification, recognition, etc.) but usually they suffer from slow convergence speed. Here, we propose to use the natural gradient instead of the classical ordinary gradient in order to enhance the convergence properties. Results are presented concerning identification and predistortion using classical and natural gradient. Practical implementations issues are given at the end of the paper. Copyright © 2002 John Wiley & Sons, Ltd. [source]

Position-Invariant Neural Network for Digital Pavement Crack Analysis

Byoung Jik Lee
This system includes three neural networks: (1) image-based neural network, (2) histogram-based neural network, and (3) proximity-based neural network. These three neural networks were developed to classify various crack types based on the subimages (crack tiles) rather than crack pixels in digital pavement images. These spatial neural networks were trained using artificially generated data following the Federal Highway Administration (FHWA) guidelines. The optimal architecture of each neural network was determined based on the testing results from different sets of the number of hidden units, learning coefficients, and the number of training epochs. To validate the system, actual pavement pictures taken from pavements as well as the computer-generated data were used. The proximity value is determined by computing relative distribution of crack tiles within the image. The proximity-based neural network effectively searches the patterns of various crack types in both horizontal and vertical directions while maintaining its position invariance. The final result indicates that the proximity-based neural network produced the best result with the accuracy of 95.2% despite its simplest neural network structure with the least computing requirement. [source]

Adaptive TS-FNN control for a class of uncertain multi-time-delay systems: The exponentially stable sliding mode-based approach

Tung-Sheng Chiang
Abstract This paper presents an adaptive Takagi,Sugeno fuzzy neural network (TS-FNN) control for a class of multiple time-delay uncertain nonlinear systems. First, we develop a sliding surface guaranteed to achieve exponential stability while considering mismatched uncertainty and unknown delays. This exponential stability result based on a novel Lyapunov,Krasovskii method is an improvement when compared with traditional schemes where only asymptotic stability is achieved. The stability analysis is transformed into a linear matrix inequalities problem independent of time delays. Then, a sliding mode control-based TS-FNN control scheme is proposed to achieve asymptotic stability for the controlled system. Since the TS-FNN combines TS fuzzy rules and a neural network structure, fewer numbers of fuzzy rules and tuning parameters are used compared with the traditional pure TS fuzzy approach. Moreover, all the fuzzy membership functions are tuned on-line even in the presence of input uncertainty. Finally, simulation results show the control performance of the proposed scheme. Copyright © 2008 John Wiley & Sons, Ltd. [source]

Memory effects description by neural networks with delayed feedback connections

Petia D. Koprinkova
For the purpose of dynamic systems modeling, it was proposed to include feedback connections or delay elements in the classical feed-forward neural network structure so that the present output of the neural network depends on its previous values. These delay elements can be connected to the hidden and/or output neurons of the main neural network. Each delay element gets a value of a state variable at a past time instant and keeps this value during a single sampling period. The groups of delay elements record the values of the state variables for a given time period in the past. Changing the number of the delay elements, which belongs to one group, a shorter or a longer time period in the past can be accounted for. Thus, the connection weights determine the influence of the past process states on the present state in a similar way as it is in the time delay kernel or cause-effect relation membership function (CER-MF) models. Specific feed-forward neural networks with time delay connections are used to solve the problem of neural network chemostat modeling as well as specific kinetic rates modeling. The weights of the feedback connections obtained during model training are discussed as the points of a time delay kernel or as the strength levels in a CER model (the points in the CER-MF). The corresponding changes in these weights with the changing time period in the past are shown. © 2004 Wiley Periodicals, Inc. [source]

On fault isolation by neural-networks-based parameter estimation techniques

EXPERT SYSTEMS, Issue 1 2007
Ramon Ferreiro Garcia
Abstract: The aim of the work is to exploit some aspects of functional approximation techniques in parameter estimation procedures applied on fault detection and isolation tasks using backpropagation neural networks as functional approximation devices. The major focus of this paper deals with the strategy used in the data selection task as applied to the determination of non-conventional process parameters, such as performance or process-efficiency indexes, which are difficult to acquire by direct measurement. The implementation and validation procedure on a real case study is carried out with the aid of the facilities supplied by commercial neural networks toolboxes, which manage databases, neural network structures and highly efficient training algorithms. [source]

Perturbation signal design for neural network based identification of multivariable nonlinear systems

Pankaj S. Kulkarni
Abstract The paper focuses on issues in experimental design for identification of nonlinear multivariable systems. Perturbation signal design is analyzed for a hybrid model structure consisting of linear and neural network structures. Input signals, designed to minimize the effects of nonlinearities during the linear model identification for the multivariable case, have been proposed and its properties have been theoretically established. The superiority of the proposed perturbation signal and the hybrid model has been demonstrated through extensive cross validations. The utility of the obtained models for control has also been proved through a case study involving MPC of a nonlinear multivariable neutralization plant. On traite dans cet article de la problématique des plans expérimentaux pour la détermination des systèmes multivariés non linéaires. La conception des signaux de perturbation est analysée pour un modèle de structure hybride composée de structures à réseaux linéaires et neuronaux. Des signaux d'entrée, con,us pour minimiser les effets des non-linéarités lors de la détermination du modèle linéaire pour le cas multivarié, sont proposés et leurs propriétés sont établies de manière théorique. La supériorité du signal de perturbation et du modèle hybride proposés est démontrée par des validations croisées poussées. L'utilité des modèles obtenus pour le contr,le est également prouvée par une étude de cas faisant intervenir le MPC d'une installation de neutralisation multivariée non linéaires. [source]