Showing papers on "Artificial neural network published in 1970"
TL;DR: It is demonstrated that identification theory implies unsupervised learning is possible in many important cases, and a general method is presented as inclusive as the one revealed here, which is effective for all the many cases wherein unsuper supervised learning is known to be possible.
Abstract: The first portion of this paper is tutorial. Beginning with a standard definition of an abstract pattern-recognition machine, "learning" is given a mathematical meaning and the distinction is made between supervised and unsupervised learning. The bibliography will help the interested reader retrace the history of learning in pattern recognition. The exposition now focuses attention on unsupervised learning. Carefully, it is explained how problems in this subject can be viewed as problems in the identification of finite mixtures, a statistical theory that has achieved some maturity. From this vantage point, it is demonstrated that identification theory implies unsupervised learning is possible in many important cases. The remaining sections present a general method for achieving unsupervised learning. Other authors have proposed schemes having greater computational convenience, but no method previously published is as inclusive as the one revealed here, which we demonstrate to be effective for all the many cases wherein unsupervised learning is known to be possible.
54 citations
31 citations
TL;DR: Among the statistical themes that are emphasized are the estimation of probabilities; the use of amounts of information and of evidence as substitutes for utility when utility is difficult to estimate; decision trees; “evolving” probabilities; and maximum, minimum, and minimax entropy in diagnosis.
Abstract: About a dozen examples are given of the use of statistical methods in research on machine intelligence, most, though not all, previously known, but not previously brought together. The topics include the application of rationality to the research as a whole; the trading of immediate gain for information; adaptive control without the identification of a model, by using smoothing techniques; phoneme recognition using distinctive features and their derivatives; the compiling of dictionaries; “botryology” or concept formation by clump finding; information retrieval; medical diagnosis; game playing and its relationship to theorem proving; design of an alphabet or of a vocabulary; and artificial neural networks. Among the statistical themes that are emphasized are the estimation of probabilities; the use of amounts of information and of evidence as substitutes for utility when utility is difficult to estimate; decision trees; “evolving” probabilities; and maximum, minimum, and minimax entropy in diagnosis. In this survey of methods it has been necessary at several points to make do with references to the literature.
24 citations
TL;DR: This work presents a novel methodology for pattern recognition that uses genetic learning to get an optimized classification system, applied to a real problem, in which it is required to distinguish three nuclear accidents that may occur in a nuclear power plant.
Abstract: This work presents a novel methodology for pattern recognition that uses genetic learning to get an optimized classification system. Each class is represented by several time series in a data base. The idea is to find clusters in the set of the training patterns of each class so that their centroids can distinguish the classes with a minimum of misclassifications. Due to the high level of difficulty found in this optimization problem and the poor prior knowledge about the patterns domain, a model based on genetic algorithm is proposed to extract this knowledge, searching for the minimum number of subclasses that leads to a maximum correctness in the classification. The goal of this model is to find how many and which are the clusters to consider. To validate the methodology, reference problems, where the best solution is wellknown, are proposed. Extending the scope of the application, the methodology is applied to a real problem, in which it is required to distinguish three nuclear accidents that may occur in a nuclear power plant. The misclassification rate was 5% in a total of 180 trials. To ratify the results an artificial neural network was designed and trained to solve the same problem. The results and comparisons are shown and commented. Transactions on Information and Communications Technologies vol 19 © 1998 WIT Press, www.witpress.com, ISSN 1743-3517
18 citations
DOI•
01 Jan 1970
TL;DR: A modern method of short-term air pollution prediction based on artificial neural networks based on the coal fired Thermal Power Plant at Sostanj shows promising results and it is expected that neural network based prediction models will be widely used in the future.
Abstract: A modern method of short-term air pollution prediction based on artificial neural networks is described. After an initial overview of neural network performance, the features of the model are described. The method is applied to the complex orography around the coal fired Thermal Power Plant at Sostanj. The results are promising and we expect that neural network based prediction models will be widely used in the future.
14 citations
DOI•
01 Jan 1970
TL;DR: A neural scheme is developed to model the rainfall-runoff behaviour of the Tiber River basin and shows that it is able to provide very accurate discharge forecasts and performances quite superior to the other two approaches.
Abstract: A neural network is developed to model the rainfall-runoff behaviour of the Tiber River basin. Performances of the neural network are then compared with the ones gained through an autoregressive model with exogenus input (ARX) and via the persistence hypothesis. The comparison shows that the neural scheme is able to provide very accurate discharge forecasts and performances quite superior to the other two approaches.
13 citations
TL;DR: Fuzzy logic, self adjustable neural networks and dynamic interaction among the input parameters of a system (instead of using net values) are among the new techniques.
Abstract: In this work classifying methods are examined from the view of Artificial Intelligence. Special reference is made to a pre-existing method of classifying rock masses (Bieniawski's classification method) and two typical attempts to use Artificial Intelligence tools are referred: a) Transference of the methodology procedure in an expert system's shell , and b) Training of a neural network with sets of inputs results in order to map the outer performance of the methodology. For an extension, machine learning is proposed as a tool for derivation of new classification methods taylored to specific systems. Fuzzy logic, self adjustable neural networks and dynamic interaction among the input parameters of a system (instead of using net values) are among the new techniques. Key-Words: Classification, Clustering, Artificial Intelligence, Expert Systems, Neural Networks, Fuzzy Logic.
11 citations
TL;DR: This work discusses how neural networks may contribute to improve the performance of robot path planners and three types of artificial neural networks are distinguished in their application to the robot path planning problem: self-organizing feature maps, optimization neural networks and pattern classification neural networks.
Abstract: Robot path planning is a problem largely addressed within the field of Artificial Intelligence (AI) since its exact optimal solution given by Computational Geometry is unfeasible to be implemented in real time. However most of classical AI methods are still unimplementable in real time. Artificial neural networks appear to be a promising paradigm to tackle this problem. We discuss how neural networks may contribute to improve the performance of robot path planners. Three types of artificial neural networks are distinguished in their application to the robot path planning problem: self-organizing feature maps, optimization neural networks and pattern classification neural networks.
9 citations
TL;DR: The proposed Neural Network -based Model Reference Adaptive Controller (NN-MRAC) can significantly improve the system behavior and force the system to follow the reference model and minimize the error between the model and plant output.
Abstract: The aim of this paper is to design a neural network based intelligent adaptive controller. It consists of an online multilayer back propagation neural network structure along with a conventional Model Reference Adaptive Control (MRAC).The training patterns for the Neural Network (NN) are obtained from the conventional PI controller. In the conventional model reference adaptive control (MRAC) scheme, the controller is designed to realize plant output converges to reference model output based on the plant which is linear. The NN is used to compensate the nonlinearity of the plant that is not taken into consideration in the conventional MRAC. The control input to the plant is given by the sum of the output of conventional MRAC and the output of NN. The proposed Neural Network -based Model Reference Adaptive Controller (NN-MRAC) can significantly improve the system behavior and force the system to follow the reference model and minimize the error between the model and plant output. The effectiveness of the proposal control scheme is demonstrated by simulations.
8 citations
TL;DR: This paper presents a study that was done with the recognition of the whisky, wine, ethanol, carbon thetracloride and methanol, using the NeuroSoluction software.
Abstract: An Artificial Nose is being made to detect the smell of some substances and the results of prototype phase 0 and phase 1 are displayed. In phase 0 and phase 1 we utilize conducting polymer sensors. A pattern recognition technique based on Multi Layer Perceptron (MLP) model of Artificial neural network (ANN) is adopted. In this paper we present a study that was done with the recognition of the whisky, wine, ethanol, carbon thetracloride and methanol. The NeuroSoluction software is used in this study.
6 citations
TL;DR: In order to deal with uncertainties that exist in modelled results, statistical analyses were employed to identify confidence intervals for predicted costs and comments are discussed on the results that were obtained when artificial neural networks were developed, trained and tested on the data supplied.
TL;DR: An algorithm for rule extraction from neural networks, based on the work by Lu et al. in 1996, is developed and this algorithm, named Modified RX, is experimentally evaluated in three different domains.
Abstract: This work deals with the efficient discovery of valuable and nonobvious information from large collections of data, using Computacional Intelligence tools. For this purpose, a . study about knowledge acquirement from supervised neural networks employed for classification problems is presented. An algorithm for rule extraction from neural networks, based on the work by Lu et al. [1] in 1996, is developed. This algorithm, named Modified RX, is experimentally evaluated in three different domains. The results are compared to those obtained by classification trees. In respect of the efficacy , one observes that the successful application of the algorithm mainly depends on the knowledge representation acquired by the conecctionist model, while the eflcciency only depends on the neural network training time.
TL;DR: Despite the limitations of the data set, the results show that neural networks have potential for use as direct classifiers of river water quality from biological field data.
Abstract: The paper describes an investigation into the use of neural networks for the direct classification of river water quality from biological data. The theoretical basis of biological monitoring is briefly explained and the most commonly used methods are discussed, together with some recently developed computer-based methods. The biological basis of the study is iully described, while it is assumed that the reader has a basic understanding of artificial neural networks. The networks used were multi-layer perceptrons with a single hidden layer of eight nodes and an output layer of five nodes, one for each of the five biological-based water quality classes adopted for use in this study. Two different input sets were tested: one based upon the states of existence, in field samples, of forty-one key biological indicator organisms; and the other based upon twelve inputs derived from principal component analysis of the field data. Fifty-three field samples, previously classified by an expert river ecologist, were used for both the training and the testing of the networks, but independence between the training and test sets was maintained using one-fold cross validation. Despite the limitations of the data set, the results show that neural networks have potential for use as direct classifiers of river water quality from biological field data.
TL;DR: It is obtained that the back propagation neural network possess the strong ability in time series prediction by comparing it with the autoregresive modeling method.
Abstract: The way for predicting vibration value and forecasting serious malfunctions for large rotating machinery using neural networks is proposed in this paper. The topological architecture of a multi-layer neural network for this purpose and the training strategies are established. It is obtained that the back propagation neural network possess the strong ability in time series prediction by comparing it with the autoregresive modeling method. With the neural network, one-day prediction of the rotor vibration value for a 200 MW turbo-generator has been carried out quite accurately. The surging state in a CO2 compressor has also been prognosed in the same way but taking multi related values of the process parameters as the input of the neural net.
TL;DR: The use of neural network for fault classification in rotating systems using signal of vibration to model process and systems from actual data and to respond in real time to the changes in the machine state.
Abstract: Vibration monitoring of components in manufacturing plants involves the collection of vibration data, and a detailed analysis of data. One of the most important characteristic of neural networks is their ability to model process and systems from actual data, and to respond in real time to the changes in the machine state. This paper discusses the use of neural network for fault classification in rotating systems using signal of vibration.
TL;DR: The performance of a fuzzy neural network edge detector is compared with the neural network and the traditional techniques such as Sobel, LoG, Gabor function, and relaxation.
Abstract: In this paper, the edge detection using fuzzy neural network is described. The input features are fuzzy sets and a learning algorithm employs fuzzified delta rule. To increase the efficiency during the training, the varied learning rate and the momentum is applied instead of fixed values. In addition, instead of pixel-based inputs, the texture-based inputs are fed into the fuzzy neural network to facilitate and determine the quality of an edge feature. Experimental results have been tested for the case of both step edges and real world images with noise. The performance of a fuzzy neural network edge detector is compared with the neural network and the traditional techniques such as Sobel, LoG, Gabor function, and relaxation.
TL;DR: It is felt that the application of neural network shows superior performance in fault diagnosis, whereas conventional techniques like spectral analysis require complete processing of an input signal to reach a diagnosis.
Abstract: A method is presented for multiple fault diagnosis by means of an Artificial Neural Network (ANN). The major advantage of using an ANN as opposed to any other technique for fault diagnosis in condition ba:-ed maintenance is that the network produces an immediate decision with minimal computation for a given input vector, whereas conventional techniques like spectral analysis require complete processing of an input signal to reach a diagnosis. The basic strategy is to train a neural network to recognize the behavior of the machine condition as well as the behavior of the possible system faults. The multi-layer feed forward network is used in this paper with back propagation learning algorithm. The network is trained by giving training examples, which have known input vector of vibration signatures and output vector of membership of possible faults. Field data of a lubricating oil pump for a residual gas compressor from a LPG recovery plant is used for training and testing the network. For diagnosis purpose, five different states are considered. The correct classification rate during training and testing is very high. On the basis of the results presented it is felt that the application of neural network shows superior performance in fault diagnosis. Transactions on Information and Communications Technologies vol 6, © 1994 WIT Press, www.witpress.com, ISSN 1743-3517
TL;DR: A hybrid Machine Learning algorithm developed at the University of South Florida is presented as a knowledge acquisition tool for developing knowledge-based systems.
Abstract: Automation of the knowledge acquisition process in building knowledgebased systems for process design is addressed through Machine Learning techniques. A hybrid Machine Learning algorithm developed at the University of South Florida is presented as a knowledge acquisition tool for developing knowledge-based systems. The learning algorithm addresses the knowledge acquisition problem by developing and maintaining the knowledge base through inductive learning from the examples. The learning algorithm named as Symbolic-Connectionist net (SCnet), overcomes the problems associated with neural and symbolic learning systems by integrating the symbolic information into a neural network representation. The learning system allows for knowledge extraction and background knowledge encoding in the form of rules. Fuzzy logic has been made use of in dealing with uncertainty in the learning domain. The description language for the learning system consists of continuous and discrete variables along with relational and fuzzy comparators. The applicability of the learning system for process design is illustrated through a complex column sequencing example. The performance of the learning system is discussed in terms of the knowledge extracted from example cases and its classification accuracy on the test cases. Transactions on Information and Communications Technologies vol 1, © 1993 WIT Press, www.witpress.com, ISSN 1743-3517
01 Jan 1970
TL;DR: A method for estimating water level at Sungai Bedup in Sarawak is presented here, which makes use of Artificial Neural Network (ANN), a new tool that is capable of modeling various nonlinear hydrological processes.
Abstract: A method for estimating water level at Sungai Bedup in Sarawak is presented here. The method makes use of Artificial Neural Network (ANN) – a new tool that is capable of modeling various nonlinear hydrological processes. ANN was chosen based on its ability to generalize patterns in imprecise or noisy and ambiguous input and output data sets. In this study, the networks were developed to forecast daily water level for Sungai Bedup station. Specially designed networks were simulated using data obtained from Drainage and Irrigation Department with MATLAB 6.5 computer software. Various training parameters were considered to achieve the best result. ANN Recurrent Network using Backpropagation algorithm was adopted for this study.
TL;DR: This communication shows a new approach to inverse problems that uses a neural network and an expert system to determine the external heat transfer given observation of the temperature history at one or more interior points.
Abstract: Inverse problems are problems of determining cause on the basis of the knowledge of their effects. The object of the inverse heat conduction problem is to determine the external heat transfer (the cause) given observation of the temperature history at one or more interior points (the effect). This communication shows a new approach to inverse problems. This approach uses a neural network and an expert system. The examples shown in this paper were computed using back propagation software (neural network) and a system based on Lukasiewicz's many valued logic (expert system). The numerical technique of neural networks evolved from the effort to model the function of the human brain and expert systems take the place of expert knowledge in several areas.
TL;DR: Results are presented showing that the neural network developed can discriminate between faults on a DC traction system and the normal operation of both camshaft and chopper controlled rolling stock.
Abstract: The protection of DC traction circuits from fault conditions is a difficult task. The paper describes the development of a neural network to be used in a multifunction protection relay. The neural network developed can discriminate between normal train operation and short-circuit conditions on the DC traction supply. The neural network provides fault protection and location. The neural network was trained and tested using both simulated and actual data. The paper presents results showing that the neural network can discriminate between faults on a DC traction system and the normal operation of both camshaft and chopper controlled rolling stock. Conclusions are given on the advantages and disadvantages of using neural networks, along with their possible use in also locating a fault.
01 Jan 1970
TL;DR: Optimization techniques are at the core of data science including data analysis and machine learning, and an understanding of basic optimization techniques and their fundamental properties provides important grounding for students, researchers, and practitioners in these areas as mentioned in this paper.
Abstract: Optimization techniques are at the core of data science, including data analysis and machine learning. An understanding of basic optimization techniques and their fundamental properties provides important grounding for students, researchers, and practitioners in these areas. This text covers the fundamentals of optimization algorithms in a compact, self-contained way, focusing on the techniques most relevant to data science. An introductory chapter demonstrates that many standard problems in data science can be formulated as optimization problems. Next, many fundamental methods in optimization are described and analyzed, including: gradient and accelerated gradient methods for unconstrained optimization of smooth (especially convex) functions; the stochastic gradient method, a workhorse algorithm in machine learning; the coordinate descent approach; several key algorithms for constrained optimization problems; algorithms for minimizing nonsmooth functions arising in data science; foundations of the analysis of nonsmooth functions and optimization duality; and the back-propagation approach, relevant to neural networks.
DOI•
01 Jan 1970
TL;DR: In this article, a method of predicting benthic community structure from environmental variables using artificial neural networks is described, where a single output from the network predicts the number of individuals of a given taxon to be found in a 5.5cm by 10cm deep core sample of lake sediment taken at the site in question.
Abstract: A method of predicting benthic community structure from environmental variables using artificial neural networks is described. The input variables represent geophysical, limnological and sedimentological characteristics of sites in Canadian waters of the Laurentian Great Lakes. A single output from the network predicts the number of individuals of a given taxon to be found in a 5.5cm by 10cm deep core sample of lake sediment taken at the site in question. Networks have been trained for four taxa; namely Oligochaeta, Porifera, Chironomidae and Pelecypoda. Three input vector sets were compared: the 28 dimension raw data set, a subset of 9 variables and a 7 dimension eigenvector set. Performance tests were carried out using a 1-fold cross validation technique, which maximises data utility whilst maintaining independence between the training and test sets. Details of the training and testing of the networks are given, together with a brief introduction to neural networks. It is concluded that artificial neural networks have potential for use in biological monitoring systems.
Journal Article•
TL;DR: Feed-forward back propagation (FFBP) with TRAINBR training function, LEARNGD adaption learning function, and SSE performance function was the final type of networks try and the minimum error and 0.96 correlation coefficient were the end results.
Abstract: Artificial neural networks (ANNs) are the result of academic investigations that use mathematical formulations to model nervous system operations. Neural networks (NNs) represent a meaningfully different approach to using computers in the workplace that is used to learn patterns and relationships in data. In this paper, compressive strength (CS) of lightweight concrete with 0, 20, 30, and 50 percent of scoria instead of sand and different watercement ratio and cement content for 288 cylindrical samples have been studied. Out of them 36 samples were randomly selected to be used for this study. The CS of these samples has been used to train ANNs for CS prediction to get the optimum value. ANNs have been formed by MATLAB software that the minimum error in data training and the maximum correlation coefficient in data were final goals. For this reasons, feed-forward back propagation (FFBP) with TRAINBR training function, LEARNGD adaption learning function, and SSE performance function was the final type of networks try. The FFBP was 3-10-1 (3 inputs, 10 neurons in hidden layer, and 1 output) that the minimum error and 0.96 correlation coefficient were the end results.
DOI•
01 Jan 1970
TL;DR: In order to enhance the estimation rate of the SEEK( Switching mode Enhanced Extended Kalman) algorithm an artificial neural network is embedded within the SeeK algorithm.
Abstract: In order to enhance the estimation rate of the SEEK( Switching mode Enhanced Extended Kalman) algorithm an artificial neural network is embedded within the SEEK algorithm. The SEEK-FIND(SEEK For Initialisation of Neural Descriptor) algorithm is efficient to estimate state variables in multi-modal nonlinear systems such as contaminant dispersion processes. When a nonlinear system has regularities and irregularities such as bifurcation the neural network is applied to estimate the regular states after leaning by using the state variable estimated from the SEEK filter as teacher signals. The SEEK filter takes over again when the system shows irregularities. The SEEK filter and the neural network are automatically switched by monitoring the maximum estimation error. The numerical demonstrations indicate the state variables are accurately estimated and the estimation speed is enhanced in the SEEK-FIND algorithm
TL;DR: Artificial neural networks and genetic algorithms for solving optimization problems in ship control systems have been proposed and genetic-neural algorithm GNA for improving a quality of solutions have been introduced.
Abstract: In this paper, artificial neural networks and genetic algorithms for solving optimization problems in ship control systems have been proposed. The gradient Hopfield ANN for linear minimization is considered. Moreover, HANN for finding local Pareto-optimal solutions in multicriteria optimization of designed navigation systems has been considered. Finally, genetic-neural algorithm GNA for improving a quality of solutions have been introduced.
01 Jan 1970
TL;DR: Details from a selected variety of works published in recent years are presented to illustrate the versatility of the Deep Learning techniques, their potential in current and future research and industry applications as well as their state-of-the-art status in vision tasks, where their efficiency is experimentally proven to near 100% accuracy.
Abstract: Deep Learning usage is spread across many fields of application. This paper presents details from a selected variety of works published in recent years to illustrate the versatility of the Deep Learning techniques, their potential in current and future research and industry applications as well as their state-of-the-art status in vision tasks, where their efficiency is experimentally proven to near 100% accuracy. The presented applications range from navigation to localization, object recognition and more advanced interactions such as grasping.
TL;DR: The learning rate for classifying parts which are produced from a machine with a varying parameter is analyzed to investigate the effect of learning rate to an unsupervised neural network when applied to an inspection process.
Abstract: In this paper, we shall investigate the effect of learning rate to an unsupervised neural network when applied to an inspection process. The network we use is the Learning by Experience (LBE)[1]. Here, we analyse the learning rate for classifying parts which are produced from a machine with a varying parameter. Experiment results using IC leadframes are included.