General Regression Neural Network

Accurate prediction of mineral grades is a fundamental step in mineral exploration and resource estimation, which plays a significant role in the economic evaluation of mining projects. Currently available methods are based either on geometrical approaches or geostatistical techniques that often considers the grade as a regionalised variable. In this paper, we propose a grade estimation technique that combines multilayer feed-forward neural network (NN) and k-nearest neighbour (kNN) models to estimate the grade distribution within a mineral deposit. The models were created by using the available geological information (lithology and alteration) as well as sample locations (easting, northing, and altitude) obtained from the drill hole data. The proposed approach explicitly maintains pattern recognition over the geological features and the chemical composition (mineral grade) of the data. Prior to the estimation of grades, rock types and alterations were predicted at unsampled locatio...

In this paper we provide an alternative approach to analyze the demand for international tourism in the Balearic Islands, Spain, by using a neural network model that incorporates time-varying conditional volatility. We consider daily air passenger arrivals to Palma de Mallorca, Ibiza and Mahon, which are located in the islands of Mallorca, Ibiza and Menorca, respectively, as a proxy for international tourism demand for the Balearic Islands. Spain is a world leader in terms of total international tourist arrivals and receipts, and Mallorca is one of the most popular destinations in Spain. For tourism management and marketing, it is essential to forecast high frequency international tourist demand accurately. As it is important to provide sensible international tourism demand forecast intervals, it is also necessary to model their variances accurately. Moreover, time-varying variances provide useful information regarding the risks associated with variations in international tourist arrivals.

The rheological properties of drilling fluids, including viscosity and yield point, are essential for the effectiveness of drilling operations. Inaccurate predictions of these parameters may lead to costly complications during the drilling operation. Among artificial intelligence (AI) methods, the general regression neural network (GRNN) approach and the fuzzy logic method possess a high speed of estimation and fewer adjustable parameters than other methods. Despite the excellent capability of these two methods, they have seldom been used to predict the rheological properties of drilling fluids. Hence, through programming in MATLAB software, the capabilities of these methods in predicting the rheological properties of drilling fluids were investigated by comparison of their predictions against experimental results. The neural network contained one input layer with three inputs (clay mass, Na2Co3 concentration, and Gum Arabic concentration), one hidden layer with 38 neurons, and one output layer with three outputs (apparent viscosity (AV), plastic viscosity (PV), and yield point (YP)). In the fuzzy logic method, the optimal value of the clustering radius was considered to be 0.1 in this research. Based on the two methods designed, the value of R (about 0.99) and RMSE (about 0.5) between predicted values and the measured values of rheological properties in training and testing data were excellent. Our findings indicate that both AI methods can be utilized to predict the rheological parameters of drilling fluids with different compositions.

In an eort to forecast daily maximum ozone concentrations, many researchers have developed daily ozone forecasting models. However, this continuing worldwide environmental problem suggests the need for more accurate models. Development of these models is dicult because the meteorological variables and photochemical reactions involved in ozone formation are complex. In this study, a neural network model for forecasting daily maximum ozone levels is developed and compared with two conventional statistical models, regression and Box±Jenkins ARIMA. The results show that the neural network model is superior to the regression and Box±Jenkins ARIMA models we tested.

Acoustic emission (AE) source localization is a powerful detection method. Time Difference Mapping (TDM) method is an effective method for detecting defects in complex structures. The core of this method is to search for a point with the minimum distance away from the verification point in the time difference database. In Traditional Time Difference Mapping (T-TDM) method and Improved Time Difference Mapping (I-TDM) method, the larger database and denser grids allow the higher localization accuracy. If the location points are not included in the database, the localization accuracy of the T-TDM and I-TDM methods will be greatly affected. To solve the above problems, a new AE source localization method, Generalized Regression Neural Network Based on Time Difference Mapping (GRNN-TDM), is proposed to improve the localization accuracy in the study. In the proposed method, the time difference data of the sensor path on all nodes in the time difference mapping are used as the training input data and the coordinates of grid nodes are used as the training output data. After continuous learning and training, the neural network model predicts its possible source location with the time difference data collected from the verification point. In this paper, the localization of AE sources with T-TDM, I-TDM and GRNN-TDM methods was studied in four composite plates with different fiber layers and an aluminum plate with holes. The localization results showed that the localization accuracy of the GRNN-TDM method was higher than that of T-TDM and I-TDM methods.

Taylor & Francis makes every effort to ensure the accuracy of all the information (the "Content") contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors make no representations or warranties whatsoever as to the accuracy, completeness, or suitability for any purpose of the Content. Any opinions and views expressed in this publication are the opinions and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of the Content should not be relied upon and should be independently verified with primary sources of information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands, costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or indirectly in connection with, in relation to or arising out of the use of the Content.

Probabilistic neural networks (PNN) and general regression neural networks (GRNN) represent knowledge by simple but interpretable models that approximate the optimal classifier or predictor in the sense of expected value of the accuracy. These models require the specification of an important smoothing parameter, which is usually chosen by cross-validation or clustering. In this article, we demonstrate the problems with the cross-validation and clustering approaches to specify the smoothing parameter, discuss the relationship between this parameter and some of the data statistics, and attempt to develop a fast approach to determine the optimal value of this parameter. Finally, through experimentation, we show that our approach, referred to as a gap-based estimation approach, is superior in speed to the compared approaches, including support vector machine, and yields good and stable accuracy.

The use of the Storm Water Management Model (SWMM) to simulate Runoff-Transport phenomenon necessitates the proper calibration of the different parameters involved in the process and the effect of these parameters on the routed hydrograph. A detailed sensitivity analysis is needed to evaluate the main parameters of the Runoff-Extended Transport Blocks to establish the most sensitive parameters affecting the Runoff-Extended Transport Simulation. This type of analysis requires tedious trial and error solutions to reach the objective calibration. Arti®cial Neural Networks (ANN) are known to have the capability of solving the inverse of many problems. The objective of this study is to examine the potential use of ANN to solve the inverse parameters needed for the calibration problem. A hypothetical test area composed of Runoff-Extended Transport Blocks was used to simulate various routed hydrographs based on changing the hydrologic±hydraulics parameters within its physical engineering ranges. The results show that ANN is a powerful instrument for directly predicting the most sensitive parameters to produce comparable routed hydrograph to the measured ones.

In this paper, an online adaptive general regression neural network (OAGRNN) is presented as a direct online speed controller for a three-phase induction motor. To keep the induction motor running at its rated speed in real-time and under a variety of load conditions, the speed error and its derivative are continuously measured and fed back to the OAGRNN controller. The OAGRNN controller provides the inverter with the control signal it needs to produce the proper frequency and voltage for the induction motor instantly. Notably, the OAGRNN controller demonstrated remarkable performance without the need for a learning mode; it was able to track the desired motor speed, starting its operation from scratch. A setup utilizing a three-phase induction motor has been developed to show the high capacity of OAGRNN for tracking the desired speed of the motor while subjected to the varied load torque. The performance of OAGRNN is examined in two phases: the MATLAB simulation and the experimental setup. Furthermore, when the OAGRNN performance is compared with that of the proportional integral (PI) controller, it demonstrates its outstanding ability and superiority for online adjustments related to the three-phase induction motor's speed control.

Artificial Neural Network (ANN)-based diagnosis of medical diseases has been taken into great consideration in recent years. In this paper, two types of ANNs are used to classify effective diagnosis of Parkinson's disease. Multi-Layer Perceptron (MLP) with back-propagation learning algorithm and Radial Basis Function (RBF) ANNs were used to differentiate between clinical variables of samples (N = 195) who were suffering from Parkinson's disease and who were not. For this purpose, Parkinson's disease data set, taken from UCI machine learning database was used. Mean squared normalized error function was used to measure the usefulness of our networks during trainings and direct performance calculations. It was observed that MLP is the best classification with 93.22% accuracy for the data set. Also, we got 86.44% accuracy in RBF classification for the same data set. This technique can assist neurologists to make their ultimate decisions without hesitation and more astutely.

This correspondence presents an investigation into the comparative performance of an active vibration control (AVC) system using a number of intelligent learning algorithms. Recursive least square (RLS), evolutionary genetic algorithms (GAs), general regression neural network (GRNN), and adaptive neuro-fuzzy inference system (ANFIS) algorithms are proposed to develop the mechanisms of an AVC system. The controller is designed on the basis of optimal vibration suppression using a plant model. A simulation platform of a flexible beam system in transverse vibration using a finite difference method is considered to demonstrate the capabilities of the AVC system using RLS, GAs, GRNN, and ANFIS. The simulation model of the AVC system is implemented, tested, and its performance is assessed for the system identification models using the proposed algorithms. Finally, a comparative performance of the algorithms in implementing the model of the AVC system is presented and discussed through a set of experiments.

The transfer of energy between two adjacent parts of rock mainly depends on its thermal conductivity. Present study supports the use of artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS) in the study of thermal conductivity along with other intrinsic properties of rock due to its increasing importance in many areas of rock engineering, agronomy and geo environmental engineering field. In recent years, considerable effort has been made to develop techniques to determine these properties. Comparative analysis is made to analyze the capabilities among six different models of ANN and ANFIS. ANN models are based on feedforward backpropagation network with training functions resilient backpropagation (RP), one step secant (OSS) and Powell-Beale restarts (CGB) and radial basis with training functions generalized regression neural network (GRNN) and more efficient design radial basis network (NEWRB). A data set of 136 has been used for training different models and 15 were used for testing purposes. A statistical analysis is made to show the consistency among them. ANFIS is proved to be the best among all the networks tried in this case with average absolute percentage error of 0.03% and regression coefficient of 1, whereas best performance shown by the FFBP (RP) with average absolute error of 2.26%. Thermal conductivity is predicted using P-wave velocity, porosity, bulk density, uniaxial compressive strength of rock as input parameters.

This study aims at predicting the effects of selected process parameters on nips stability and number of nips by using different artificial intelligence methods. Partially oriented polyester yarn with 283 dtex linear density and different numbers of filaments are intermingled with different speed and pressure levels. The feed forward neural network with multi-hidden layers (ML-FFNN) and general regression neural networks (GRNN) have been selected as artificial intelligence methods. The number of filaments, intermingling speed and pressure values are used as input variables on the artificial neural networks. The effects of number of hidden layers on the ML-FFNN and number of nodes in the hidden layer are investigated. Based on comparative results, the ML-FFNN is found to give better performance (at most 6%) than by GRNN in terms of prediction accuracy on train and test data sets. It can be concluded from this study that the neural networks has great ability to predict intermingling p...

This paper presents the findings of laboratory model testing of arched bridge constrictions in a rectangular open channel flume whose bed slope was fixed at zero. Four different types of arched bridge models, namely single opening semi-circular arch (SOSC), multiple opening semi-circular arch (MOSC), single opening elliptic arch (SOE), and multiple opening elliptic arch (MOE), were used in the testing program. The normal crossing (/ = 0), and five different skew angles (/ = 10°, 20°, 30°, 40°, and 50°) were tested for each type of arched bridge model. The main aim of this study is to develop a suitable model for estimating backwater through arched bridge constrictions with normal and skewed crossings. Therefore, different artificial neural network approaches, namely multi-layer perceptron (MLP), radial basis neural network (RBNN), generalized regression neural network (GRNN), and multi-linear and multi-nonlinear regression models, MLR and MNLR, respectively were used. Results of these experimental studies were compared with those obtained by the MLP, RBNN, GRNN, MLR, and MNLR approaches. The MLP produced more accurate predictions than those of the others.

Unmanned air vehicles (UAV's) today are extensively used for a wide range of applications, from amateur to human to military applications. Electric propulsion is preferred for small UAV's, while piston engines and gas turbines are used for the bigger ones. The proposed engine in this work covers the midscale for restart able, low budget vehicles (<10kg and 30-200N thrust). Simplicity, durability and manufacturability are the main target goals of the design. The liquid fuelled engine is designed to run on gasoline and ethanol as a fuel and a hydrogen peroxide -water solution as the oxidizer. The advantage of the high energy density of the liquid reactants is counteracted by the difficulty for the ignition of the reactants at low pressures and temperatures, especially at the starting up process. In this work, different nozzles and injection strategies were tested and evaluated under similar conditions that exist at start up of the engine. It has been found that even for reasonable manufacturing process and precision, good atomization can be obtained if a mixture of gas and liquid is premixed prior to injection. This can be realized by catalytic decomposition of the hydrogen peroxide without the need for an extra gas supply circuit. This exothermal reaction provides the additional advantage of injecting a hot mixture in the cold combustion chamber, allowing easier ignition.

The transfer of energy between two adjacent parts of rock mainly depends on its thermal conductivity. Present study supports the use of artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS) in the study of thermal conductivity along with other intrinsic properties of rock due to its increasing importance in many areas of rock engineering, agronomy and geo environmental engineering field. In recent years, considerable effort has been made to develop techniques to determine these properties. Comparative analysis is made to analyze the capabilities among six different models of ANN and ANFIS. ANN models are based on feedforward backpropagation network with training functions resilient backpropagation (RP), one step secant (OSS) and Powell-Beale restarts (CGB) and radial basis with training functions generalized regression neural network (GRNN) and more efficient design radial basis network (NEWRB). A data set of 136 has been used for training different models and 15 were used for testing purposes. A statistical analysis is made to show the consistency among them. ANFIS is proved to be the best among all the networks tried in this case with average absolute percentage error of 0.03% and regression coefficient of 1, whereas best performance shown by the FFBP (RP) with average absolute error of 2.26%. Thermal conductivity is predicted using P-wave velocity, porosity, bulk density, uniaxial compressive strength of rock as input parameters.

In the study of crystalline materials, the lattice constant (LC) of perovskites compounds play important role in the identification of materials. It reveals various interesting properties. In this study, we have employed Support Vector Regression, Artificial Neural Network, and Generalized Regression Neural Network based Computational Intelligent (CI) techniques to predict LC of cubic and monoclinic perovskites. Due to their interesting physiochemical properties, investigations in modeling the structural properties of perovskites have gained considerable attention. A dataset of a reasonable number of cubic and monoclinic perovskites are collected from the current literature. The CI techniques can efficiently correlate the LC of the perovskites materials with the ionic radii of constituent elements. A performance analysis of CI techniques is carried out with Multiple Linear Regression techniques, SPuDS software, and Density-Functional Theory. We have observed that the CI techniques yield accurate LC prediction as against the conventional approaches. Availability: Matlab based computer program developed for this work is available on request.

When combining a set of learned models to form an improved estimator, the issue of redundancy or multicollinearity in the set of models must be addressed. A progression of existing approaches and their limitations with respect to the redundancy is discussed. A new approach, PCR*, based on principal components regression is proposed to address these limitations. An evaluation of the new approach on a collection of domains reveals that: 1) PCR* was the most robust combination method as the redundancy of the learned models increased, 2) redundancy could be handled without eliminating any of the learned models, and 3) the principal components of the learned models provided a continuum of \regularized" weights from which PCR* could choose.

Neural decoding is an important task for understanding how the biological nervous system performs computation and communication. This paper introduces a novel continuous neural decoding method based on general regression neural network (GRNN). GRNN does not require an iterative training procedure as in other neural networks and allows the appropriate regression form to be expressed as a probability density function (pdf) that is empirically determined from the observed data using nonparametric estimators with no underlying assumptions. These characteristics make the GRNN decoder appropriate for continuous neural decoding problem. We compared the performance of the GRNN model with commonly used Wiener filter model in decoding rats' neural activity of Motor Cortex (M1) during rats' lever pressing task. Experiments show that GRNN has superior capacity for neural decoding.

Purpose: The goal of this paper is to develop a pragmatic system of a production throughput forecasting system for an automated test operation in a hard drive manufacturing plant. The accurate forecasting result is necessary for the management team to response to any changes in the production processes and the resources allocations.Design/methodology/approach: In this study, we design a production throughput forecasting system in an automated test operation in hard drive manufacturing plant. In the proposed system, consists of three main stages. In the first stage, a mutual information method was adopted for selecting the relevant inputs into the forecasting model. In the second stage, a generalized regression neural network (GRNN) was implemented in the forecasting model development phase. Finally, forecasting accuracy was improved by searching the optimal smoothing parameter which selected from comparisons result among three optimization algorithms: particle swarm optimization (PS...

Single speech feature such as Mel-Frequency Cepstral Coefficient (MFCC) has been used in most of the studies to classify asphyxia cry among infants. Other speech features such as Chromagram, Mel-scaled Spectrogram, Spectral Contrast and Tonnetz have not been reported in any study related to the classification of asphyxia cry. The study investigated the use of hybrid features of MFCC, Chromagram, Mel-scaled Spectrogram, Spectral Contrast and Tonnetz and deep learning models in classifying asphyxia cry. Deep learning models such as Deep Neural Network (DNN) and Convolutional Neural Network (CNN) were used to classify infant cry between normal/non-asphyxia and asphyxia. The performance of the deep learning models was compared using concatenated hybrid features and single feature of MFCC. The Baby Chillanto Database was used in this study. CNN model performed better than DNN models when MFCC was used. DNN models performed better with hybrid features compared to that with single feature of MFCC. DNN with multiple hidden layers achieved an accuracy of 100% in classifying normal and asphyxia cry, and 99.96% for non-asphyxia and asphyxia cry when the hybrid features were used.

This paper discusses the application of General Regression Neural Network (GRNN) for predicting the software quality attributefault ratio. This study is carried out using static Object-Oriented (OO) measures (64 in total) as the independent variables and fault ratio as the dependent variable. Software metrics used include those concerning inheritance, size, cohesion and coupling. Prediction models are designed using 15 possible combinations of the four categories of the measures. We also tested the goodness of fit of the neural network model with the standard parameters. Our study is conducted in an academic institution with the software developed by students of Undergraduate/Graduate courses.

Multiscale Image Processing and Artificial Neural Networks (ANNs). In mammography diagnosis, Generalized Regression Neural Network (GRNN) is a succesful novelty. A number of relevant features are extracted from Regions of Interest (ROI) to reduce the classifier complexity. We compared our results with a Feed Forward Neural Networks (FFNNs). Our performance results successfully compete with classical FFNNs and present a much faster training phase, overcoming also minima problem.

Analysis of reliability data plays an important role in the maintenance decision making process. The accurate estimation of residual life in components and systems can be a great asset when planning the preventive replacement of components on machines. Artificial intelligence is a field that has rapidly developed over the last twenty years and practical applications have been found in many diverse areas. The use of such methods in the maintenance field have however not yet been fully explored.

The intelligent acoustic emission locator is described in Part I, while Part II discusses blind source separation, time delay estimation and location of two simultaneously active continuous acoustic emission sources. The location of acoustic emission on complicated aircraft frame structures is a difficult problem of non-destructive testing. This article describes an intelligent acoustic emission source locator. The intelligent locator comprises a sensor antenna and a general regression neural network, which solves the location problem based on learning from examples. Locator performance was tested on different test specimens. Tests have shown that the accuracy of location depends on sound velocity and attenuation in the specimen, the dimensions of the tested area, and the properties of stored data. The location accuracy achieved by the intelligent locator is comparable to that obtained by the conventional triangulation method, while the applicability of the intelligent locator is more general since analysis of sonic ray paths is avoided. This is a promising method for non-destructive testing of aircraft frame structures by the acoustic emission method.

The Springer Geophysics series seeks to publish a broad portfolio of scientific books, aiming at researchers, students, and everyone interested in geophysics. The series includes peer-reviewed monographs, edited volumes, textbooks, and conference proceedings. It covers the entire research area including, but not limited to, applied geophysics, computational geophysics, electrical and electromagnetic geophysics, geodesy, geodynamics, geomagnetism, gravity, lithosphere research, paleomagnetism, planetology, tectonophysics, thermal geophysics, and seismology.

Ventricular Tachyarrhythmias, especially Ventricular Fibrillation, are the primary arrhythmias which are cause of sudden death. The object of this study is to characterize Ventricular Fibrillation prior to its onset because only care is in-time defibrillation. Two prediction methodologies are being presented here i.e. Neural Network and Nearest Neighbour. Standard Nearest Neighbour method is modified and evaluated for Prediction. Electrocardiograph signals of patients are studied having three types of arrhythmia i.e. Ventricular Tachycardia, Ventricular Flutter and Ventricular Fibrillation. Electrocardiographs of subjects having normal sinus rhythm are also studied. For Classification of theses signals, Generalized Regression Neural Network, standard and modified Nearest Neighbour methods used. VF is often proceeds by episodes of VT. These methods can recognize VT class so onset of VF can be predicted before time. Promising results are found.

Complex variability is a significant problem in predicting construction crew productivity. Neural Networks using supervised learning methods like Feed Forward Back Propagation (FFBP) and General Regression Neural Networks (GRNN) have proved to be more advantageous than statistical methods like multiple regression, considering factors like the modelling ease and the prediction accuracy. Neural Networks using unsupervised learning like Self Organizing Maps (SOM) have additionally been introduced as methods that overcome some of the weaknesses of supervised learning methods through their clustering ability. The objective of this article is thus to compare the performances of FFBP, GRNN and SOM in predicting construction crew productivity. Related data has been collected from 117 plastering crews through a systematized time study and comparison of prediction performances of the three methods showed that SOM have a superior performance in predicting plastering crew productivity.

Advance knowledge of which files in the next release of a large software system are most likely to contain the largest numbers of faults can be a very valuable asset. To accomplish this, a negative binomial regression model has been developed and used to predict the expected number of faults in each file of the next release of a system. The predictions are based on the code of the file in the current release, and fault and modification history of the file from previous releases. The model has been applied two large industrial systems, one with a history of 17 consecutive quarterly releases over 4 years, and the other with nine releases over 2 years. The predictions were quite accurate: for each release of the two systems, the 20 percent of the files with the highest predicted number of faults contained between 71 percent and 92 percent of the faults that were actually detected, with the overall average being 83 percent. The same model was also used to predict which files of the firs...

The colloquium speaker will discussed a methodology that utilizes a generalized regression neural network to develop a hybrid option trading system that incorporates both volatility and return forecasting. This study focuses on the S&P 500 stock index as being representative of the market. Two different hybrid systems are discussed. The first hybrid system applies a signal from the volatility forecasting as a primary signal, and then uses long and short straddle, strip, and strap strategies to take advantage of the volatility signal. The second hybrid system applies a signal from the return forecasting as a primary signal, and then uses long calls and puts, and bull and bear spread strategies to take advantage of the forecasting signal. Results show that the hybrid options trading model can improve the overall trading return and can outperform trading models using merely return forecasting or volatility forecasting in isolation.

The forecasting of exchange rates remains a difficult task due to global crises and authority interventions. This study employs the monetary-portfolio balance exchange rate model and its unrestricted version in the analysis of Malaysian Ringgit during the post-Bretton Wood era (1991M1-2012M12), before and after the subprime crisis. We compare two Artificial Neural Network (ANN) estimation procedures (MLFN and GRNN) with the random walks (RW) and the Vector Autoregressive (VAR) methods. The out-ofsample forecasting assessment reveals the following. First, the unrestricted model has superior forecasting performance compared to the original model during the 24-month forecasting horizon. Second, the ANNs have outperformed both the RW and VAR forecasts in all cases. Third, the MLFNs consistently outperform the GRNNs in both exchange rate models in all evaluation criteria. Fourth, forecasting performance is weakened when the post-subprime crisis period was included. In brief, economic fundamentals are still vital in forecasting the Malaysian Ringgit, but the monetary mechanism may not sufficiently work through foreign exchange adjustments in the short run due to global uncertainties. These findings are beneficial for policy making, investment modelling, and corporate planning.

In Part I, an intelligent acoustic emission (AE) locator is described while the Part II discusses a blind source separation, time delay estimation and location of two continuous AE sources. AE analysis is used for characterization and location of developing defects in materials. AE sources often generate a mixture of various statistically independent signals. A difficult problem of AE analysis is a separation and characterization of signal components when the signals from various sources and the mode of mixing are unknown. Recently, blind source separation (BSS) by independent component analysis (ICA) has been used to solve these problems. The purpose of this paper is to demonstrate the applicability of ICA to locate two independent simultaneously active AE sources on an aluminum band specimen. The method is promising for non-destructive testing of aircraft frame structures by AE analysis.

Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public.

The transfer of energy between two adjacent parts of rock mainly depends on its thermal conductivity. Present study supports the use of artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS) in the study of thermal conductivity along with other intrinsic properties of rock due to its increasing importance in many areas of rock engineering, agronomy and geo environmental engineering field. In recent years, considerable effort has been made to develop techniques to determine these properties. Comparative analysis is made to analyze the capabilities among six different models of ANN and ANFIS. ANN models are based on feedforward backpropagation network with training functions resilient backpropagation (RP), one step secant (OSS) and Powell-Beale restarts (CGB) and radial basis with training functions generalized regression neural network (GRNN) and more efficient design radial basis network (NEWRB). A data set of 136 has been used for training different models and 15 were used for testing purposes. A statistical analysis is made to show the consistency among them. ANFIS is proved to be the best among all the networks tried in this case with average absolute percentage error of 0.03% and regression coefficient of 1, whereas best performance shown by the FFBP (RP) with average absolute error of 2.26%. Thermal conductivity is predicted using P-wave velocity, porosity, bulk density, uniaxial compressive strength of rock as input parameters.

Complete supervised training algorithms for B-spline neurJJ networks and fuzzy rule-based systems are discu.~•;ed. By introducing the relationships between B-spline neural networks and certain types of l'uzzy models, training algorithms developed initially for neural networks can be adapted to fuzzy systems. The standard training criterion is reformulated, by separating its parameters into linear and non linear ones. By employing this reformulation with the Levenberg-Marquardt algorithm, a new training method. olfering a last rate of convergence, and a robust perlormance compared with the standard Error-Back Propagation algorithm. Copyright t) 20021FAC

There is no doubt that so-called "artificial neural networks" (ANN) are powerful computational tools to model complex nonlinear systems. In my view, an ANN establishes a data-driven nonlinear relationship between inputs and outputs of a system. The fact that such a nonlinear model is generally very complicated (usually one does not even write down the equations) renders it a black-box model. The fact that the model contains numerous parameters makes imperative the use of an advanced optimization method to calibrate its parameters. Once an ANN is fitted, it can be used to predict outputs from known inputs. Thus, there have been numerous successful applications of ANN in forecasting the future evolution of complex systems (e.g. Casdagli & Eubank, 1992; Weigend & Gershenfeld, 1994). However, I am afraid that there has also been an abuse in other cases, indirectly assisted by the numerous technical details, inapproachable for the majority of scientists (in our case hydrologists), and even by the exotic ANN vocabulary. The paper by Kişi (2006) stimulated my "reflex" questions about "neural networks", their use and abuse, and helped me to organize them so that they can be addressed to the "central nervous system". I start with the vocabulary itself. How "neural" are ANN? This question may be philosophical and related to the nature of the so called "artificial intelligence". I am not prepared to discuss in depth this type of question, for which profound analyses can be found in Dreyfus & Dreyfus (1986), Penrose (1989, 1994, 1996) and Hodges (2000-2002). However, I thought of a simplified version of this question: Is it necessary to call these mathematical models "neural" networks? Structures such as graphs and networks are very common in science (including hydrology, hydraulics and water resources), and consist of nodes (or vertices) and arcs (edges, or arrows in directed graphs and networks). Why in this case should we speak of "neurons" instead of nodes or arcs? After all, while in other cases (e.g. a river network, a water distribution network) the network has a physical hypostasis, in ANN the network is just a convenient pictorial representation of a complicated mathematical nonlinear relationship, which could be handled even without the notion of a network. A lot of similar questions can arise: Why use such terms as "one-pass learning algorithm with a highly parallel structure" or "training vector"? Isn't it more understandable to speak about "model calibration", "model fitting" and "parameter optimization" instead of "learning" or "training"? If yes, why have these anthropomorphized terms been so prevalent? Furthermore, is an ANN approach appropriate for any type of problem? Kişi (2006) seems to reply positively to this question. His motivation for the paper is simply this: "However, the application of ANN to evapotranspiration modelling is limited in the literature. ... To the knowledge of the author, no work has been reported

There is no doubt that so-called "artificial neural networks" (ANN) are powerful computational tools to model complex nonlinear systems. In my view, an ANN establishes a data-driven nonlinear relationship between inputs and outputs of a system. The fact that such a nonlinear model is generally very complicated (usually one does not even write down the equations) renders it a black-box model. The fact that the model contains numerous parameters makes imperative the use of an advanced optimization method to calibrate its parameters. Once an ANN is fitted, it can be used to predict outputs from known inputs. Thus, there have been numerous successful applications of ANN in forecasting the future evolution of complex systems (e.g. Casdagli & Eubank, 1992; Weigend & Gershenfeld, 1994). However, I am afraid that there has also been an abuse in other cases, indirectly assisted by the numerous technical details, inapproachable for the majority of scientists (in our case hydrologists), and even by the exotic ANN vocabulary. The paper by Kişi (2006) stimulated my "reflex" questions about "neural networks", their use and abuse, and helped me to organize them so that they can be addressed to the "central nervous system". I start with the vocabulary itself. How "neural" are ANN? This question may be philosophical and related to the nature of the so called "artificial intelligence". I am not prepared to discuss in depth this type of question, for which profound analyses can be found in Dreyfus & Dreyfus (1986), Penrose (1989, 1994, 1996) and Hodges (2000-2002). However, I thought of a simplified version of this question: Is it necessary to call these mathematical models "neural" networks? Structures such as graphs and networks are very common in science (including hydrology, hydraulics and water resources), and consist of nodes (or vertices) and arcs (edges, or arrows in directed graphs and networks). Why in this case should we speak of "neurons" instead of nodes or arcs? After all, while in other cases (e.g. a river network, a water distribution network) the network has a physical hypostasis, in ANN the network is just a convenient pictorial representation of a complicated mathematical nonlinear relationship, which could be handled even without the notion of a network. A lot of similar questions can arise: Why use such terms as "one-pass learning algorithm with a highly parallel structure" or "training vector"? Isn't it more understandable to speak about "model calibration", "model fitting" and "parameter optimization" instead of "learning" or "training"? If yes, why have these anthropomorphized terms been so prevalent? Furthermore, is an ANN approach appropriate for any type of problem? Kişi (2006) seems to reply positively to this question. His motivation for the paper is simply this: "However, the application of ANN to evapotranspiration modelling is limited in the literature. ... To the knowledge of the author, no work has been reported

The transfer of energy between two adjacent parts of rock mainly depends on its thermal conductivity. Present study supports the use of artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS) in the study of thermal conductivity along with other intrinsic properties of rock due to its increasing importance in many areas of rock engineering, agronomy and geo environmental engineering field. In recent years, considerable effort has been made to develop techniques to determine these properties. Comparative analysis is made to analyze the capabilities among six different models of ANN and ANFIS. ANN models are based on feedforward backpropagation network with training functions resilient backpropagation (RP), one step secant (OSS) and Powell-Beale restarts (CGB) and radial basis with training functions generalized regression neural network (GRNN) and more efficient design radial basis network (NEWRB). A data set of 136 has been used for training different models and 15 were used for testing purposes. A statistical analysis is made to show the consistency among them. ANFIS is proved to be the best among all the networks tried in this case with average absolute percentage error of 0.03% and regression coefficient of 1, whereas best performance shown by the FFBP (RP) with average absolute error of 2.26%. Thermal conductivity is predicted using P-wave velocity, porosity, bulk density, uniaxial compressive strength of rock as input parameters.

TEZ11731Tez (Doktora) -- Çukurova Üniversitesi, Adana, 2019.Kaynakça (s. 91-95) var.xviii, 97 s. : res. (bzs. rnk.), tablo ; 29 cm.Derin öğrenme yapıları birçok bilimsel çalışma alanında önceden görülmemiş başarı oranları yakalamıştır. En çok bilinen derin öğrenme yapısı olan konvolüsyonel sinir ağı genel olarak örüntü tanıma çalışmalarında kullanılmaktadır. Konvolüsyonel sinir ağ yapıları konvolüsyon ve havuzlama katmanları içeren bir özellik çıkarıcı yapısı ve tam bağlı bir sınıflayıcı ağ yapısından oluşur. Bu tezin ilk çalışmasında, güçlü bir çekirdek tabanlı sınıflayıcı olan GCNN, sınıflandırma performansını arttırmak için CNN yapısına adapte edilmiştir. Bu adaptasyon kesinlik, duyarlılık ve F1-skor değerlerinde sırası ile %44.45, %39.69 ve %43.57 değerlerine kadar bağıl bir performans artışı sağlamıştır. Bu adaptasyon performansı önemli ölçüde arttırmasına karşın, konvolüsyonel kısmın temsil yeteneği anlamında zayıf kaldığı gözlemlenmiştir. Bu nedenle, daha yüksek bir öğrenme p...

A tunable radial basis function (RBF) network model is proposed for nonlinear system identification using particle swarm optimisation (PSO). At each stage of orthogonal forward regression (OFR) model construction, PSO optimises one RBF unit's centre vector and diagonal covariance matrix by minimising the leave-one-out (LOO) mean square error (MSE). This PSO aided OFR automatically determines how many tunable RBF nodes are sufficient for modelling. Compared with the-state-of-the-art local regularisation assisted orthogonal least squares algorithm based on the LOO MSE criterion for constructing fixed-node RBF network models, the PSO tuned RBF model construction produces more parsimonious RBF models with better generalisation performance and is computationally more efficient.

Transient stability evaluation (TSE) is part of dynamic security assessment of power systems, which involves the evaluation of the system's ability to remain in equilibrium under credible contingencies. Neural networks (NN) have been applied to the security assessment of power systems and have shown great potential for predicting the security of power systems. This paper proposes a generalized regression neural networks (GRNN) based classification for transient stability evaluation in power systems. In the proposed method, learning data sets have been generated using time domain simulation (TDS). The GRNN input nodes representing the voltage magnitude for all buses, real and reactive powers on transmission lines, the output node representing the transient stability index. The proposed GRNN was implemented and tested on IEEE 9-bus and 39-bus test systems. NN results show that the stability condition of the power system can be predicted with high accuracy and less misclassification rate.

Transient stability evaluation (TSE) is part of dynamic security assessment of power systems, which involves the evaluation of the system's ability to remain in equilibrium under credible contingencies. Neural networks (NN) have been applied to the security assessment of power systems and have shown great potential for predicting the security of power systems. This paper proposes a generalized regression neural networks (GRNN) based classification for transient stability evaluation in power systems. In the proposed method, learning data sets have been generated using time domain simulation (TDS). The GRNN input nodes representing the voltage magnitude for all buses, real and reactive powers on transmission lines, the output node representing the transient stability index. The proposed GRNN was implemented and tested on IEEE 9-bus and 39-bus test systems. NN results show that the stability condition of the power system can be predicted with high accuracy and less misclassification rate.

Magnetic resonance imaging (MRI) logs are well logs that use nuclear magnetic resonance to accurately measure free fluid, irreducible water (MBVI), and effective porosity (MPHI). Permeability is then calculated using a mathematical function that incorporates these measured properties. This paper describes the methodology developed to generate synthetic magnetic resonance imaging logs using data obtained by conventional well logs such as spontaneous potential (SP), gamma-ray, caliper, and resistivity. The synthetically generated logs are named virtual magnetic resonance logs or 'VMRL' for short. Magnetic resonance logs provide the capability of in-situ measurement of reservoir characteristics. The study also examines and provides alternatives for situations in which all required conventional logs are unavailable for a particular well. Synthetic magnetic resonance logs for wells with an incomplete suite of conventional logs are generated and compared with actual magnetic resonance logs for the same well. In order to demonstrate the feasibility of the concept being introduced here, the methodology is applied in two different fashions. First, it is applied to four wells; each from a different part of the country. These wells are located in East Texas, Gulf of Mexico, Utah, and New Mexico. Since only one well from each region is available, the model is developed using a segment of the pay zone and consequently is applied to the rest of the pay zone. In a second approach, the technique is applied to a set of wells in a highly heterogeneous reservoir in East Texas. Here the model was developed using a set of wells and then was verified by applying it to a well away from the wells used during the development process. This technique is capable of providing a better reservoir description (effective porosity, fluid saturation, and permeability) and more realistic reserve estimation at a much lower cost.

MRI logs are well logs that use nuclear magnetic resonance to accurately measure free fluid, irreducible water (MBVI), and effective porosity (MPHI). Permeability is then calculated using a mathematical function that incorporates these measured properties. This paper describes the methodology developed to generate synthetic Magnetic Resonance Imaging logs using data obtained by conventional well logs such as SP, Gamma Ray, Caliper, and Resistivity. The synthetically generated logs are named Virtual Magnetic Imaging Logs or &quot;VMRI&quot; logs for short. This methodology incorporates artificial neural networks as its main tool. Virtual MRI logs for irreducible water saturation (MBVI) and effective porosity (MPHI) as well as permeability (MPERM) were generated for four wells. These wells are located in East Texas, Gulf of Mexico, Utah, and New Mexico. The results are quite encouraging. It is shown that MPHI, MBVI, and MPERM logs can be generated with a high degree of accuracy. For e...

One of the costliest parts of field-scale reservoir studies is log analysis. A recent GRI project required a detailed study of a field with hundreds of wells. As part of this study all the well logs were to be analyzed by an engineer in order to identify net pay, porosity, and saturation. It soon became apparent that a considerable amount of time must be devoted to well log analysis in order to obtain consistent and high quality reservoir characteristics throughout the field. This was mainly due to the fact that logs for several wells were missing and many wells did not have the suite of logs that were necessary for analysis. Virtual intelligence techniques are used in construction of the intelligent software tool presented in this study.

Accurate demand forecasting is one of the most key issues in inventory management of spare parts. The problem of modeling future consumption becomes especially difficult for lumpy patterns, which characterized by intervals in which there is no demand and, periods with actual demand occurrences with large variation in demand levels. However, many of the forecasting methods may perform poorly when demand for an item is lumpy. In this study based on the characteristic of lumpy demand patterns of spare parts a hybrid forecasting approach has been developed, which use a multi-layered perceptron neural network and a traditional recursive method for forecasting future demands. In the described approach the multi-layered perceptron are adapted to forecast occurrences of non-zero demands, and then a conventional recursive method is used to estimate the quantity of non-zero demands. In order to evaluate the performance of the proposed approach, their forecasts were compared to those obtained ...

This paper compares one-step-ahead out-of-sample predictions on Malaysian Ringgit-US Dollar exchange rate using the generalized regression neural network for a range of forecasting horizons from 1991M3 to 2008M8. We find that the monetary fundamentals are significant in explaining the dynamics of Malaysian exchange rate in a longer forecast horizon as the performance of monetary exchange rate models outperformed the random walk benchmark model. The results also revealed that Malaysian exchange rate market provides profitable short-term arbitrage opportunities with lagged observations, and the integration of autoregressive terms into the monetary exchange rate models enhanced the out-of-sample forecasting performance.

The transfer of energy between two adjacent parts of rock mainly depends on its thermal conductivity. Present study supports the use of artificial neural network (ANN) and adaptive neuro fuzzy inference system (ANFIS) in the study of thermal conductivity along with other intrinsic properties of rock due to its increasing importance in many areas of rock engineering, agronomy and geo environmental engineering field. In recent years, considerable effort has been made to develop techniques to determine these properties. Comparative analysis is made to analyze the capabilities among six different models of ANN and ANFIS. ANN models are based on feedforward backpropagation network with training functions resilient backpropagation (RP), one step secant (OSS) and Powell-Beale restarts (CGB) and radial basis with training functions generalized regression neural network (GRNN) and more efficient design radial basis network (NEWRB). A data set of 136 has been used for training different models and 15 were used for testing purposes. A statistical analysis is made to show the consistency among them. ANFIS is proved to be the best among all the networks tried in this case with average absolute percentage error of 0.03% and regression coefficient of 1, whereas best performance shown by the FFBP (RP) with average absolute error of 2.26%. Thermal conductivity is predicted using P-wave velocity, porosity, bulk density, uniaxial compressive strength of rock as input parameters.