International Journal of Neural Networks and Advanced Applications
E-ISSN: 2313-0563
Volume 1, 2014
Notice: As of 2014 and for the forthcoming years, the publication frequency/periodicity of NAUN Journals is adapted to the 'continuously updated' model. What this means is that instead of being separated into issues, new papers will be added on a continuous basis, allowing a more regular flow and shorter publication times. The papers will appear in reverse order, therefore the most recent one will be on top.
Submit a paper | Submission terms | Paper format
Volume 1, 2014
Title of the Paper: Design Features and Research on the Neuro-Like Learning Control System of a Vehicle
Authors: I. Kobersy, V. Finaev, D. Beloglazov, I. Shapovalov, J. Zargaryan, V. Soloviev
Pages: 73-80
Abstract: The application features of neuro-like learning control systems of vehicles are considered in this paper. The structure of a fuzzy control system, the description of a system operation are given, and the necessity of design of the control system integrating advantages both fuzzy and neuron systems are substantiated Here we analyzed the operation of well-known hybrid control systems of mobile robots and developed the fuzzy control system of the vehicle. The fuzzy control system of the vehicle has a modular structure. The modules of the fuzzy control system are neuro-fuzzy networks. We used genetic algorithms for training the modules. Research of training algorithms was conducted. Information about the duration of search for optimal parameters of the system modules and accuracy of obtained results is presented. We developed the special application for researches.
Title of the Paper: Fast Information Retrieval from Big Data by Using Neural Networks Implemented in the Frequency Domain
Authors: Hazem M. El-Bakry, Nikos E. Mastorakis, Michael E. Fafalios
Pages: 68-72
Abstract: The objective of storing data is to retrieve it as requested in a fast way. In this paper, a new efficient model for fast retrieving of specific information from big data is presented. Fast neural networks are used to find the best matching between words in query and stored big data. The idea is to accelerate the searching operation in a big data. This is done by applying cross correlation between the given query and the big data in the frequency domain rather than time domain. Furthermore, neural networks are used to retrieve information from big data even these data are noised or distorted. The mathematical prove for the acceleration process is analyzed and a formula for the theoretical speed up ratio is given. Simulation results confirm the theoretical considerations.
Title of the Paper: Assessment of Large Scale Urban Mapping from Airborne Hyperspectral Data based on SVM and ANN
Authors: Lamyaa Gamal El-Deen Taha, Attia Abd Al Fattah Shahin
Pages: 56-67
Abstract: Hyperspectral imaging has tremendous application such areas as mineral identification, crop classification, land use planning, environment monitoring, and military reconnaissance. In this research the potential of airborne hyperspectral data for large scale urban mapping was evaluated in order to know to what scale of airborne hyperspectral data is suitable for map production. Geometric correction using direct georefrencing and atmospheric correction have been performed. After that feature selection has been applied. Quality of rectification has been assessed using DGPS check points. A program has been developed using paython language for assessment of geopositioning accuracy of airborne hyperspectral data. The results revealed that the total RMS of airborne hyperspectral data was 0.4 m. The RMS was within the required standard map accuracy for a map 1:1000 according to the map accuracy standards of NMAS. Planimetric vector map has been produced manually and automatically. For automatic map production a comparison between two subpixel classifiers (Support vector machines (SVM) and artificial neural network (ANN)) was carried out. Quantitative accuracy assessments of the classification results were performed. Experiment shows that, Support vector machine technique outperformed neural network technique in terms of the overall classification accuracy and kappa coefficient. The overall accuracy of the Support vector machine method was 98%, and kappa coefficient was 0.96 and the overall accuracy of the neural network method was 96%, and kappa coefficient was 0.94. After that morphological operations were performed in order to remove noise followed by raster to vector conversion. The results of the SVM and the results of the neural network classification were compared and showed an increase in accuracy of land use discrimination using SVM. In conclusion by analysis of the results, it is obvious that the manual map production method is faster than the automatic methods and automatic methods needs manual editing compared to the resulted vector map from on screen digitizing.
Title of the Paper: Efficient Weather Forecasting using Artificial Neural Network as Function Approximator
Authors: I. El-Feghi, Z. Zubia, S. Abozgaya
Pages: 49-55
Abstract: Forecasting is the referred to as the process of estimation in unknown situations. Weather forecasting, especially air temperature, is one of the most important factors in many applications. This paper presents an approach to develop Artificial Neural Networks (ANNs) to forecast air temperatures. One important architecture of neural networks namely Radial Basis Function (RBF) will be used as a function approximator. The RBF trained using meteorological data of one year and tested on another year. The data consist of observations of various meteorological variables such as relative humidity, dew point, wind speed, wind direction and air pressure. To come up with appropriate centers for the RBF neurons, the weather data was clustered into several groups using kmeans clustering algorithm. The goal of developing this network is to forecast the air temperature and to have a regression model with minimum error of prediction The data of one year is used for supervised training using labelled data while the other year data was used of testing the trained ANN. Several testes were run to come up with the most suitable ANN structure based on lowest MSE in predicting air temperature. Results have shown that these structures gave very good prediction in term of accuracy.
Title of the Paper: Contribution to the Artificial Neural Network Direct Control of Torque Application Utilizing Double Stars Induction Motor
Authors: Hechelef Mohammed, Abdelkader Meroufel, Abdelrahman Guebli
Pages: 43-48
Abstract: In this paper we propose to study a control strategy known as Neural Network direct control for the study of the control of torque when using speed loop regulation of double start induction motor. The research discussed below indicates that it is possible to replace a conventional switching table by a neural network. The neural networks used are the back-propagation, to reduce the training patterns and increase the execution speed of the training process, the inputs of switching table are converted to digital signals, i.e., one bit represent the flux error, one bit the torque error, and one bit represent the location of tension Vector. As results we achieved can be summarised as follows: 1-amelioration the responding time of the system 2-Minimization of the torque ripples. 3-Minimization of the current total harmonic distortion
Title of the Paper: Stability Evaluation of Neural and Statistical Classifiers based on Modified Semi-bounded Plug-in Algorithm
Authors: Ibtissem Ben Othman, Faouzi Ghorbel
Pages: 37-42
Abstract: This paper illustrates a new criterion for evaluating neural networks stability compared to the Bayesian classifier. The stability comparison is performed by the error rate probability densities estimation using the modified semi-bounded Plug-in algorithm. We attempt, in this work, to demonstrate that the Bayesian approach for neural networks improves the performance and stability degree of the classical neural classifiers.
Title of the Paper: RF Coverage Analysis and Validation of Cellular Mobile Data using Neural Network
Authors: Z. Nadir, M. Bait-Suwailam, M. Shafiq
Pages: 30-36
Abstract: This paper provides an extension of pathloss analysis in Urban environments in Oman. The paper addresses the applicability of Okumura-Hata model in an area in Oman in GSM frequency band of 890-960 MHz. The root mean square error (RMSE) was calculated between measured Pathloss values and those predicated on the basis of Okumura-Hata model for that area. Artificial Neural Network (ANN) was also used to forecast the data for a much larger distance. ANN provides a wide and rich class of reliable and powerful statistical tools to mimic complex nonlinear functional relationships. The networks are then trained by learning through empirical data. These trained neural nets are finally used to make desired forecasts. These results are acceptable and can be used for OMAN.
Title of the Paper: Image Authentication with Tampering Localization using Chaotic and Neural Mapping
Authors: Dattatherya, K. Suresh, M. MadhaviLatha, Manoj Kumar Singh
Pages: 20-29
Abstract: This paper proposes a new approach to authenticate the image using combination of different chaotic maps along with a newly developed a kind of chaotic neural network .They are integrated to develop the authentication code for each pixel available in the image. This authentication code is carried by transmitted image itself near lossless image content quality and need not required any extra memory requirement. Initial parameter sensitivity and aperiodicity of chaotic maps and its integration with other chaotic map and a chaos based neural network are utilized to create the uncertainty in predictability of authentication code in association with image pixels. Developed method is applicable to work for gray as well color images with very little extra requirement in terms of computational complexity. Proposed method is not only authenticated the image but also define the locations of tampering hidden in the image with very high efficiency.
Title of the Paper: A New Algorithm of Neural Internal Model Controller using Variable Learning Rate
Authors: Ayachi Errachdi, Mohamed Benrejeb
Pages: 13-19
Abstract: In this paper is proposed a new algorithm of internal model controller using a proposed variable learning rate. This method overcomes the difficulty to choose a fixed parameter (0 1) when we apply an Adaptive method for Internal Model based on neural network Control (AIMC) for nonlinear systems. This work clarify that if the learning rate is large ( 1) , learning may occur quickly, but it may also become unstable or if the learning rate is small ( 0) learning adapt reliably, but it may take a long time and thus, it can invalidate the purpose of real-time operation. The proposed method is dependent on the availability of the inverse neural model of the system and the availability of the internal neural model in each moment. These two neural models use the proposed variable learning. The bloc of on-line inverse model and on-line internal model will be used as a bloc of neural controller. This bloc of controller tries to minimize the error between the system output and the internal model output. The adjustment of the controller bloc runs in each moment. The robustness of the proposed adaptive internal model neural network control strategy is investigated in threes cases; firstly when the system has time-invariant parameters, secondly when it has a time-varying parameters and finally when it’s a noisy time-varying system. The proposed strategy is compared with the Adaptive Direct Inverse Control (ADIC). From the experiments, it is showing that the performance of the AIMC method is much better than the ADIC method. Two different reference command signals are used to test the control system performance, and it is noted that an excellent tracking response is exhibited in the presence of disturbance.
Title of the Paper: Modeling the Flashover Voltage Using ANN
Authors: Kherfane Riad Lakhdar, Younes Mimoun, Khodja Fouad, Kherfane Naas
Pages: 7-12
Abstract: This work attempts to apply an artificial intelligent technique which is the ANN to estimate the flashover voltage for polluted insulators, based on various studies published in this field and given by reference to the experimental results on insulators artificially polluted. The obtained results are promising and insure that ANN technique can help researchers in this field to understand more deeply and estimate the critical flashover voltage for new designed insulators.
Title of the Paper: A New Algorithm for MRAC Method using a Neural Variable Learning Rate
Authors: Ayachi Errachdi, Mohamed Benrejeb
Pages: 1-6
Abstract: This paper presents a new algorithm for MRAC (Model Reference Adaptive Control) method based-on neural networks using a variable learning rate. The proposed mechanism adaptation algorithm demonstrates that if the learning rate is large, learning may occur quickly, but it may also become unstable or if the learning rate is small learning adapt reliably, but it may take a long time and thus, it can invalidate the purpose of real-time operation. To overcome these problems we propose a neural controller using variable learning rate. This corresponding algorithm depends on the error between the actual plant output and the output of the reference model. The control strategy is based on two-steps; the first is initialization parameters of the neural controller using reduced number of observation. In the second phase, the parameters of the neural controller are directly tuned from the training data via the tracking error. The simulation results show that the proposed algorithm using variable learning rate is simple to implement and may be extended to multivariable system.
