Search results

Presents recollections of the scientists who were the pioneers of cybernetics, together with an insight into their personalities and lives. Discusses their individual contributions as the basis for such currently important fields as Artificial Intelligence, Neural Nets, Computer Technology and Robotics. Show these distinguished men as being deeply and vitally involved with larger human and ethical issues.

Combining the parallel and neural paradigms seems, at first glance, to be a natural process, since it is a methodology derived from the part played by the biological and mathematical behavior of a neuron. It is proposed that any neural algorithm is inherently a parallel application. The structure of a neural algorithm and the function of a neuron suggest the choice of the systolic approach. However, interest should be restricted only to those well‐known neural models such as the Hopfield and back‐propagation neural networks. It is also shown that the systolic approach is best suited to the parallelization of the patterns training phase of the neural algorithms in terms of mapping the two structures (systolic and neural networks).

Biological oscillatory activity in neural networks has been intensively studied over the past years. Neuronal oscillations are the basis of many different behavioral patterns and sensory mechanism. Understanding the dynamic properties of these mechanisms is useful for analyses of biological functions and medical diagnoses. The dynamic characteristics of wake‐sleep circadian rhythms and ECG’s cardiac cycle data measured for normal subjects are identified here, using MA‐BP neural network model. It was found that dynamics of regular components can be captured by the model. The captured dynamics are kept in a steady state for some periods. The order of the MA neural network was suppressively controlled by the first 2∼3 orders. Hence it may be useful for medical diagnoses of circadian rhythms and heart related diseases.

To develop a new predictive control scheme based on neural networks for linear and non‐linear dynamical systems.

The approach relies on three different multilayer neural networks using input‐output information with delays. One NN is used to identify the process under control, the other is used to predict the future values of the control error and finally the third one is used to compute the magnitude of the control input to be applied to the plant.

This scheme has been tested by controlling discrete‐time SISO and MIMO processes already known in the control literature and the results have been compared with other control approaches with no predictive effects. Transient behavior of the new algorithm, as well as the steady state one, are observed and analyzed in each case studied. Also, online and offline neural network training are compared for the proposed scheme.

The theoretical proof of stability of the proposed scheme still remains to be studied. Conditions under which non‐linear plants together with the proposed controller present a stable behavior have to be derived.

The main advantage of the proposed method is that the predictive effect allows to suitable control complex non‐linear process, eliminating oscillations during the transient response. This will be useful for control engineers to control complex industrial plants.

This general approach is based on predicting the future control errors through a predictive neural network, taking advantage of the NN characteristics to approximate any kind of relationship. The advantage of this predictive scheme is that the knowledge of the future reference values is not needed, since the information used to train the predictive NN is based on present and past values of the control error. Since the plant parameters are unknown, the identification NN is used to back‐propagate the control error from the output of the plant to the output of the controller. The weights of the controller NN are adjusted so that the present and future values of the control error are minimized.

Radial basis function (RBF) neural networks form an essential category of architectures of neurocomputing. They exhibit interesting and useful properties of stable and fast learning associated with significant generalization capabilities. This successful performance of RBF neural networks can be attributed to the use of a collection of properly selected RBFs. In this way this category of the networks strongly relies on some domain knowledge about a classification problem at hand. Following this vein, this study introduces fuzzy clustering, and fussy isodata, in particular, as an efficient tool aimed at constructing receptive fields of RBF neural networks. It is shown that the functions describing these fields are completely derived as a by‐product of fuzzy clustering and do not require any further tedious refinements. The efficiency of the design is illustrated with the use of synthetic two‐dimensional data as well as real‐world highly dimensional ECG patterns. The classification of the latter data set clearly points out advantages of RBF neural networks in pattern recognition problems.

This paper details how the complexity can be reduced in conic section function neural network (CSFNN) by using sensitivity analysis and the results are given for various problems. This is, particularly important for neural network hardware applications. The method used here extracts the cause and effect relationship between the inputs and outputs of the network. After training a neural network, one may want to know the effect that each of the network inputs is having on the network output. The input channels that produce low sensitivity values can be considered insignificant and can most often be removed from the network. This will reduce the size of the network, which in turn reduces the complexity and the training time.

The purpose of this paper is to develop a system to analyse the characteristics of infrared objects.

According to the gray scale of image pixel by the image entropy, gray scale estimating is carries on to construct the neural networks. And then the grey relational analysis and grey clustering methods are applied to filter the possible object. The target is predicted through image segmentation pretreatment based on the forecasting value by grey system and assigned corresponding mark. The forecasting precision is greatly elevated by GM (1, 1) model.

The paper illustrates that, based on the analysis and its experimental results, this system has a good recognition rate for infrared target.

This paper provides a way to grasp the minutial feature of the image. The filtering operation based on pixel level provided auto‐adapted filtering with a new stage.

Applications of grey theory deepened the content of detecting infrared targets and enriched technology of image processing.

This system introduces an effective method for detecting infrared targets.

The area of artificial neural networks, which dates back to the early twentieth century, could only offer positive contributions to technology after the back‐propagation algorithm was proposed in 1986. In this note an alternative algorithm to the gradient descent used in back‐propagation is proposed. This algorithm is based on the discrete central difference. This procedure, as opposed to the back‐propagation algorithm, offers the possibility of true parallel computation.

The purpose of this paper is to define a new method (grey relational analysis (GRA)) for extracting pattern samples of dissolved gases in power transformer oil, then a hybrid algorithm of the back‐propagation (BP) network and fuzzy genetic algorithm‐artificial neural network (FGA‐ANN) is used to power transformer fault diagnosis based on extracted pattern samples.

The existing manners (e.g. international electro technical commission triple‐ratio method), in practice, have certain faultiness due to the ambiguity of the inference and insufficient standard for judgment. So GRA method is chosen to solve a problem of optimal pattern samples data, then a hybrid algorithm of the BP network and FGA‐ANN is developed to optimize initial weights and to enable fast convergence of the BP network, and lastly, this algorithm is applied to the classification of dissolved gas analysis (DGA) data and power transformer fault diagnosis.

If possible, the results should be accompanied by significance. For comparative studies, the proposed scheme does not require the three ratio code and high diagnosis accuracy is obtained. In addition, useful information is provided for future fault trends and multiple faults analysis.

Accessibility and availability of data are the main limitations which model will be applied.

This paper provides useful advice for power transformer fault diagnosis method based on DGA data.

The new method of optimal choice of options of pattern samples due to GRA. The paper is aimed at optimized samples data classified and abandons the traditional ratio method.