Back-propagation Multilayer Perceptron Research Articles

Chaotic time series prediction based on brain emotional learning model and self-adaptive genetic algorithm

Chaos phenomenon is one of the most important physical phenomena, which has significant effects on one's production and life. Therefore, it is indispensable to find out the regularity of chaotic time series from a chaotic system for weather forecasting, space missions, alarm systems, etc. Although various models and learning algorithms have been developed to predict chaotic time series, many traditional methods suffer drawbacks of high computational complexity, slow convergence speed, and low prediction accuracy, due to extremely complex dynamic characteristics of chaotic systems. In this paper, a brain-inspired prediction model, i.e., brain emotional learning (BEL) model combined with self-adaptive genetic algorithm (AGA) is proposed. The establishment of BEL model is inspired by the neurobiology research, which has been put forward by mimicking the high-speed emotional learning mechanism between amygdala and orbitofrontal cortex in mammalian brain, it has advantages of lowcomputational complexity and fast learning. The BEL model employs reward-based reinforcement learning to adjust the weights of amygdala and orbitofrontal cortex. However, the reward-based method is modelsensitive and hard to generalize to other issues. To improve the performance of BEL model, AGA-BEL is proposed for chaotic prediction, in which the AGA is employed for parameter optimization. Firstly, weights and biases of orbitofrontal cortex and amygdala in BEL model are distributed to chromosomal gene sequence for optimization. Secondly, fitness function is employed to adjust the weights of amygdale and orbitofrontal cortex by self-adaptive crossover and mutation operations Therefore, the parameter optimization problem is transformed into a function optimization problem in the search space. Finally, the best chromosome that represents the best combination of weights and biases for BEL model is chosen, which is used for chaotic prediction. Prediction experiments on the benchmark Lorenz chaotic time series and a real-world chaotic time series of geomagnetic activity Dst index are performed. The experimental results and numerical analysis show that the proposed AGA-BEL prediction model achieves lower mean absolute deviation, mean square error, mean absolute percentage error, and higher correlation coefficient than the original BEL, levenberg marquardt-back propagation (LM-BP) and multilayer perceptron-back propagation (MLP-BP). Meanwhile, the BEL-based models take less computational time than the traditional BP-based models. Therefore, the proposed AGA-BEL model possesses the advantages of fast learning and low computational complexity of BEL model as well as the global optimum solution of AGA. It is superior to other traditional methods in terms of prediction precision, execution speed, and stability, and it is suited for online prediction in fast-varying environments.

Prediction of ultimate strength of shale using artificial neural network

A rock failure criterion is very important for prediction of the ultimate strength in rock mechanics and geotechnics; it is determined for rock mechanics studies in mining, civil, and oil wellborn drilling operations. Also shales are among the most difficult to treat formations. Therefore, in this research work, using the artificial neural network (ANN), a model was built to predict the ultimate strength of shale, and comparison was made with support vector machine (SVM), multiple linear regression models, and the widely used conventional polyaxial failure criteria in the stability analysis of rock structures, Drucker-Prager, and Mogi-Coulomb. For building the model, the corresponding results of triaxial and polyaxial tests have been performed on shales by various researchers. They were collected from reliable published articles. The results obtained showed that a feed forward back propagation multi-layer perceptron (MLP) was used and trained using the Levenberg–Marquardt algorithm, and the 2-4-1 architecture with root-mean-square-error (RMSE) of 24.41 exhibits a better performance in predicting the ultimate strength of shale in comparison with the investigated models. Also for further validation, triaxial tests were performed on the deep shale specimens. They were prepared from the Ramshire oilfield in SW Iran. The results obtained were compared with ANN, SVM, multiple linear regression models, and the conventional failure criterion prediction. They showed that the ANN model predicted ultimate strength with a minimum error and RMSE being equal to 43.81. Then the model was used for prediction of the threshold broken pressure shale layer in the Gachsaran oilfield in Iran. For this, a vertical and horizontal stress was calculated based on a depth of shale layer. The threshold broken pressure was calculated for the beginning and ending of a shale layer to be 154.21 and 167.98 Mpa, respectively.

Wavelet denoising of partial discharge signals and their pattern classification using artificial neural networks and support vector machines

Este artículo presenta dos enfoques de reconocimiento de patrones usando huellas dactilares de descargas parciales como características de entrada para llevar a cabo la clasificación de patrones de DP. Un perceptrón multicapa (MLP) basado en el algoritmo de propagación hacia atrás y una máquina de soporte vectorial fueron entrenados para reconocer tres tipos de patrones de DP. Los resultados experimentales demostraron que los algoritmos pueden arrojar altas tasas de reconocimiento. De otra parte, la trasformada wavelet discreta (DWT) fue utilizada para eliminar el nivel de ruido presente en las DP como una etapa previa al proceso de clasificación. Diferentes wavelets madre fueron probadas a diferentes niveles de descomposición con el objeto de encontrar parámetros wavelet apropiados para obtener una mejor relación señal-ruido (SNR) y menos distorsión después del proceso de filtrado.

Convective heat transfer and pressure drop study on nanofluids in double-walled reactor by developing an optimal multilayer perceptron artificial neural network

The effect of nanofluid on the cooling performance and pressure drop of a jacked reactor has experimentally been investigated. Aqueous nanofluids of Al2O3 and CuO was used as the cool ant inside the cooling jacket of the reactor. The application of the artificial neural networks (ANNs) to predict the performance of a double-walled reactor has been studied. Different architectures of artificial neural networks were developed to predict the convective heat transfer and pressure drop of nanofluids. The experimental results are used for training and testing the ANNs based on two optimal models via feed-forward back-propagation multilayer perceptron (MLP). The comparison of statistical criteria of different network shows that the optimal structure for predicting the convective heat transfer coefficient is the MLP network with one hidden layer and 10 neurons, which has been trained with Levenberg–Marquardt (LM) algorithm. The predicted pressure drop values by the MLP network with two hidden layers and 6 neurons in the each layer has been used from LM training algorithm, which showed a reasonable agreement with the experimental results.

Prediction of axial capacity of piles driven in non-cohesive soils based on neural networks approach

This paper presents an application of two advanced approaches, Artificial Neural Networks (ANN) and Princi­pal Component Analysis (PCA) in predicting the axial pile capacity. The combination of these two approaches allowed the development of an ANN model that provides more accurate axial capacity predictions. The model makes use of Back-Propagation Multi-Layer Perceptron (BPMLP) with Bayesian Regularization (BR), and it is established through the incorporation of approximately 415 data sets obtained from data published in the literature for a wide range of un-cemented soils and pile configurations. The compiled database includes, respectively 247 and 168 loading tests on large-and low-displacement driven piles. The contributions of the soil above and below pile toe to the pile base resistance are pre-evaluated using separate finite element (FE) analyses. The assessment of the predictive performance of the new method against a number of traditional SPT-based approaches indicates that the developed model has attractive capabili­ties and advantages that render it a promising tool. To facilitate its use, the developed model is translated into simple design equations based on statistical approaches.

Fast multilayer perceptron neural network‐based control algorithm for shunt compensator in distribution systems

In this study, a fast learning method of back-propagation (BP) multilayer perceptron neural network-based control algorithm for shunt compensator in three-phase distribution systems is presented. The proposed method comprises of quadratic linear and non-linear errors to determine optimisation criterion error function to train the BP algorithm while the existing methods have used only linear quadratic error term. The newly developed optimisation criterion error function accelerates the convergence efficiency of BP algorithm for performance improvement of shunt compensator at point of common coupling under non-linear loading conditions. With the help of the proposed algorithm, the weighted amplitude of fundamental active and reactive current components of the load current are extracted from which the reference source currents are estimated. The performance analysis of the proposed algorithm has been evaluated using two case studies for zero-voltage regulation and power factor correction. The total harmonic distortions are improved in comparison with standard BP algorithm which has been validated in above-mentioned two case studies. This is the quite important advantage of the proposed control algorithm to improve the power quality over existing control algorithms for shunt compensator.

Spectral entropy feature subset selection using SEPCOR to detect alcoholic impact on gamma sub band visual event related potentials of multichannel electroencephalograms (EEG)

The problem of analyzing and identifying regions of high discrimination between alcoholics and controls in a multichannel electroencephalogram (EEG) signal is modeled as a feature subset selection technique that can improve the recognition rate between both groups. Several studies have reported efficient detection of alcoholics by feature extraction and selection in gamma band visual event related potentials (ERP) of a multichannel EEG signal. However, in these studies the correlation between features and their class information is not considered for feature selection. This may lead to redundancy in the feature set and result in over fitting. Therefore in this study, a statistical feature selection technique based on Separability & Correlation analysis (SEPCOR) is proposed to select an optimal feature subset automatically that possesses minimum correlation between selected channels and maximum class separation. The optimal feature selection consists of a ranking method that assigns ranks to channels based on a variability measure (V-measure). From the ranked feature set of highly discriminative features, different subsets are automatically selected by heuristically applying a correlation threshold in steps from 0.02 to 0.1. These subsets are applied as input features to multilayer perceptron (MLP) neural network and k-nearest neighbor (k-NN) classifiers to discriminate alcoholic and control visual ERP. Prior to feature selection, spectral entropy features are computed in gamma sub band (30–55Hz) interval of a 61-channel multi-trial EEG signal with multiple object recognition tasks. Independent Component Analysis (ICA) is performed on raw EEG data to remove eye blink, motion and muscle artifacts. Results indicate that both classifiers exhibit excellent classification accuracy of 99.6%, for a feature subset of 22 optimal channels with correlation threshold of 0.1. In terms of computation time, k-NN classifier outperforms multilayer perceptron-back propagation (MLP-BP) network with 7.93s whereas MLP network takes 55s to perform the recognition task with the same accuracy. Compared to feature section methods used in previous studies on the same EEG alcoholic database, there is a significant improvement in classification accuracy based on the proposed SEPCOR method.

Accurate prediction of nanofluid viscosity using a multilayer perceptron artificial neural network (MLP-ANN)

Viscosity is a significant physical property of nanofluids in practical heat transfer applications. No general model is capable for accurate prediction of nanofluid viscosity in a broad range of effective parameters. In the present work, 1490 experimental data points on relative viscosity of different nanofluids have been collected by comprehensive literature search. Then, a feed-forward back-propagation multilayer perceptron artificial neural network was developed and tested via employing Levenberg–Marquardt training algorithm in order to predict nanofluid viscosity in broad ranges of operating parameters. The model input parameters are temperature, nanoparticle size, density, volume fraction, and base fluid viscosity. Regression statistical analysis results considering training and test data (R2=0.99998), comparison of the ANN predicted values with corresponding experimental data (average absolute relative deviation (AARD)=0.41%, and maximum average relative deviation (ARD)=6.44 %) and some available theoretical models have revealed high prediction capability of the developed neural network.

Classification and characterization of blueberry mechanical damage with time evolution using reflectance, transmittance and interactance imaging spectroscopy

The aim of this work was to evaluate the performance of hyperspectral data coupled with chemometrics methods in characterizing and detecting the non-visible mechanical damage of blueberries with time evolution. Reflectance and transmittance as well as interactance hypercubes were automatically segmented by the region growing based algorithms. The maximum-normalized spectra were pretreated by the Standard Normal Variate algorithm, and subsequently the Competitive Adaptive Reweighted Sampling algorithm was applied to extract the damage-specific wavelengths. Based on confusion matrices and area under Receiver Operating Characteristics curves, transmittance showed relatively superior performance to reflectance and interactance. Application of new sample set subjected to impact tests with time evolution, results demonstrated that it was especially difficult to distinguish fresh damage in blueberry. At 2days after impacted, several transmittance-based classifiers obtained satisfactory accuracies for classifying damaged (sound) blueberries: logistic regression 79.1% (85.7%), multilayer perceptron-back propagation 74.4% (92.1%) and logistic function tree 72.1% (95.2%). Furthermore, the physical property preliminarily proved to be more pronounced than the absorbed impact energy for damage incidence and severity of blueberry via the use of multiple comparison.

Suspect screening of large numbers of emerging contaminants in environmental waters using artificial neural networks for chromatographic retention time prediction and high resolution mass spectrometry data analysis

The recent development of broad-scope high resolution mass spectrometry (HRMS) screening methods has resulted in a much improved capability for new compound identification in environmental samples. However, positive identifications at the ng/L concentration level rely on analytical reference standards for chromatographic retention time (tR) and mass spectral comparisons. Chromatographic tR prediction can play a role in increasing confidence in suspect screening efforts for new compounds in the environment, especially when standards are not available, but reliable methods are lacking. The current work focuses on the development of artificial neural networks (ANNs) for tR prediction in gradient reversed-phase liquid chromatography and applied along with HRMS data to suspect screening of wastewater and environmental surface water samples. Based on a compound tR dataset of >500 compounds, an optimized 4-layer back-propagation multi-layer perceptron model enabled predictions for 85% of all compounds to within 2min of their measured tR for training (n=344) and verification (n=100) datasets. To evaluate the ANN ability for generalization to new data, the model was further tested using 100 randomly selected compounds and revealed 95% prediction accuracy within the 2-minute elution interval. Given the increasing concern on the presence of drug metabolites and other transformation products (TPs) in the aquatic environment, the model was applied along with HRMS data for preliminary identification of pharmaceutically-related compounds in real samples. Examples of compounds where reference standards were subsequently acquired and later confirmed are also presented. To our knowledge, this work presents for the first time, the successful application of an accurate retention time predictor and HRMS data-mining using the largest number of compounds to preliminarily identify new or emerging contaminants in wastewater and surface waters.

An Efficient Face Recognition with ANN using Hybrid Feature Extraction Methods

a day's face recognition is the most interesting and active research area in the field of phycology, neuroscience, and computer vision. In this paper a fast efficient algorithm is developed with the better recognition rate of face in different conditions that is illumination, head pose, expressions etc. In which we have extract the global and local features using PCA (Principle Component Analysis) and LBP (Local Binary Pattern) respectively. we have experiment the proposed algorithm with the standard Yale database which contains 15 individuals with each contains 11 images(15*11). So the fusion of Global and Local features are fed to the MLP (Multilayer Perceptron). The BPMLP (Backpropagation Multilayer Perceptron) is used for the classification. The proposed method achieves 93% accuracy compare to the existing approach.

Modeling of CO2 loading in aqueous solutions of piperazine: Application of an enhanced artificial neural network algorithm

Carbon dioxide (CO2) separation by chemical absorption is widely regarded as the most effective method for its mitigation in natural gas streams or flue gas of the fossil fuel power plants. In this paper, CO2 loading (αCO2) in aqueous solution of piperazine (PZ: C4H10N2), a high reactive cyclic secondary diamine, is modeled over extensive ranges of operational conditions: temperature (287–395 K), PZ concentration (0.1–8 mol kg H2O−1) and CO2 partial pressure (0.0215–9510.3 kPa). To achieve this goal, a feed-forward back-propagation multilayer perceptron artificial neural network (FFANN) was developed and tested via employing the Levenberg–Marquardt training algorithm, enhanced through combination with Bayesian regularization technique. Regression analysis results (R2 = 0.9977) and comparison of the ANN predicted αCO2 values with corresponding experimental data (%AARD = 2.3927) as well as with some correlations in the literature, have revealed high prediction ability and robustness of the developed neural network.

Plant leaf recognition using texture and shape features with neural classifiers

This paper proposes a novel methodology of characterizing and recognizing plant leaves using a combination of texture and shape features. Texture of the leaf is modeled using Gabor filter and gray level co-occurrence matrix (GLCM) while shape of the leaf is captured using a set of curvelet transform coefficients together with invariant moments. Since these features are in general sensitive to the orientation and scaling of the leaf image, a pre-processing stage prior to feature extraction is applied to make corrections for varying translation, rotation and scaling factors. Efficacy of the proposed methods is studied by using two neural classifiers: a neuro-fuzzy controller (NFC) and a feed-forward back-propagation multi-layered perceptron (MLP) to discriminate between 31 classes of leaves. The features have been applied individually as well as in combination to investigate how recognition accuracies can be improved. Experimental results demonstrate that the proposed approach is effective in recognizing leaves with varying texture, shape, size and orientations to an acceptable degree.

Fragrance Measurement of Scented Rice using Electronic Nose

Abstract This article describes about an instrument and method for aroma based quality detection of Basmati and other aromatic rice varieties. It comprises few modules such as odour delivery module, sniffing module, water bath module and computing module. Odour handling module helps to deliver odour to the sensor array; a sniffing unit comprising a sensor array module that includes a eight number of metal oxide semiconductor sensors assembled on a printed circuit board, said printed circuit board fitted into a sensor chamber; a water bath module for preparing rice sample, said water bath module including a heater attachment to facilitate cooking; a computing module to quantify the aroma data acquired by sensors; data acquisition module etc. Principal Component Analysis (PCA) implemented for clustering the data sets acquired from sensor array. Also data generated from sensor array was fed to Probabilistic Neural Network (PNN), Back-propagation Multilayer Perceptron (BPMLP) and Linear Discriminant Analysis (LDA) for identification of different rice varieties. Finally, for aroma quantifying, pure-quadratic response surface methodology model used with mean square error (MSE) 0.0028.

A quantum backpropagation multilayer perceptron (QBMLP) for predicting iron adsorption capacity of calcareous soil from aqueous solution

Removal of miscible hazardous materials from aqueous solutions is an alarming problem for the environmental scientists. Several linear and nonlinear regression models like Langmuir, Freundlich, D–R, Tempkin isotherm models are in vogue for determining the adsorbing capacity of standard adsorbents used for this purpose.In this article, we propose a novel quantum inspired backpropagation multilayer perceptron (QBMLP) based on quantum gates (single qubit rotation gates and two qubit controlled-not gates) for the prediction of this adsorption behavior exhibited by calcareous soil oftentimes used in adsorbing miscible iron from aqueous solutions. The backpropagation learning formulae for the proposed QBMLP architecture has also been generalized for multiple number of layers in both field homogeneous and field heterogeneous configurations characterized by three standard activations, viz., sigmoid, tanh and tan1.5h functions.Applications of the efficiency of the proposed QBMLP over the regression models are demonstrated with regards to the prediction behavior of the adsorption of iron by calcareous soil from an aqueous solution with effect to various characteristic adsorbent parameters. The adsorption process is considered to be a physical one since the activation energy (EA) of ferrous ion adsorption is 9.469kJmol−1 due to Arrhenius. Moreover, the thermodynamic parameters of Gibb's free energy (G0), enthalpy (H0) and entropy (S0) values indicate it be spontaneous.Results indicate that QBMLP predicts the adsorption behavior of calcareous soil to a very closer extent thereby obviating the need for further regression/experimental analysis. Comparison with the performance of a similar classical multilayer perceptron (MLP) architecture also reveals the prediction and time efficiency of the proposed QBMLP architecture.

Neural networks for predicting breeding values and genetic gains

Analysis using Artificial Neural Networks has been described as an approach in the decision-making process that, although incipient, has been reported as presenting high potential for use in animal and plant breeding. In this study, we introduce the procedure of using the expanded data set for training the network. Wealso proposed using statistical parameters to estimate the breeding value of genotypes in simulated scenarios, in addition to the mean phenotypic value in a feed-forward back propagation multilayer perceptron network. After evaluating artificial neural network configurations, our results showed its superiority to estimates based on linear models, as well as its applicability in the genetic value prediction process. The results further indicated the good generalization performance of the neural network model in several additional validation experiments.

Ocean surface Stokes drift from scatterometer observations

Stokes drift, a mean Lagrangian motion induced by ocean surface gravity waves, plays an important role in modifying the upper ocean circulation. In this study, two empirical models are developed to estimate ocean surface Stokes drift Uss, based on a multilayer-perceptron Back-Propagation Neural Network (BP NN) algorithm, using buoy data collocated with European Remote Sensing (ERS) data, and using (1) bulk wave parameters and (2) spectral wave density. Both BP NN models perform reasonably well, with correlation coefficients of 0.92 between the retrieved Uss and buoy measurements, for the former bulk-parameter-based NN model, compared to 0.87, for the latter spectral wave-density-based NN model. Uss values were estimated over the global ocean during 27–29 August 1998. We found Uss in the range from 2 to 36 cm s−1 with an average of 9 cm s−1, which is in good agreement with previous observations. With the retrieved Uss, we calculated the Langmuir number Lat and the scaled Langmuir vertical turbulent velocity. The distribution of Lat (generally in the range 0.25–0.58) provides a measure of the dominance of Langmuir turbulence over kinetic energy turbulence in regions with high Uss, such as the westerlies in the Southern Ocean.

A Tutorial on Different Classification Techniques for Remotely Sensed Imagery Datasets

Classification techniques are used on large databases to develop models describing different data classes. Such analysis can provide deep insight for better understanding of different large-scale databases. Studies related to knowledge acquisition and effective knowledge development are also very popular in the remote sensing field with satellite imagery datasets. In any remote sensing research, the decision-making process mainly depends on the effectiveness of the classification process. Efficient classification techniques were developed and applied to the Statlog (Landsat Satellite) database at the University of California, Irvine Machine Learning Repository to identify six land type classes. We used three different classification algorithms on the large satellite imagery: multilayer perceptron backpropagation neural network (MLP BPNN), support vector machine (SVM), and k-nearest neighbor (k-NN). This research study aimed to evaluate the performance of these classification algorithms in the prediction of the classified lands from this large set of satellite imagery. We used different performance measures, such as classification accuracy, root-mean-square error, kappa statistic, true positive rate, false positive rate, precision, recall, and F-measure to evaluate the performance of each classifier. Among the three classification techniques applied, MLP BPNN turned out to be the best; next was k-NN, followed by SVM.

An ANN-based approach to predict blast-induced ground vibration of Gol-E-Gohar iron ore mine, Iran

Blast-induced ground vibration is one of the inevitable outcomes of blasting in mining projects and may cause substantial damage to rock mass as well as nearby structures and human beings. In this paper, an attempt has been made to present an application of artificial neural network (ANN) to predict the blast-induced ground vibration of the Gol-E-Gohar (GEG) iron mine, Iran. A four-layer feed-forward back propagation multi-layer perceptron (MLP) was used and trained with Levenberg–Marquardt algorithm. To construct ANN models, the maximum charge per delay, distance from blasting face to monitoring point, stemming and hole depth were taken as inputs, whereas peak particle velocity (PPV) was considered as an output parameter. A database consisting of 69 data sets recorded at strategic and vulnerable locations of GEG iron mine was used to train and test the generalization capability of ANN models. Coefficient of determination (R2) and mean square error (MSE) were chosen as the indicators of the performance of the networks. A network with architecture 4-11-5-1 and R2 of 0.957 and MSE of 0.000722 was found to be optimum. To demonstrate the supremacy of ANN approach, the same 69 data sets were used for the prediction of PPV with four common empirical models as well as multiple linear regression (MLR) analysis. The results revealed that the proposed ANN approach performs better than empirical and MLR models.

Prediction of Deflection and Stresses of Laminated Composite Plate with an Artificial Neural Network Aid

This paper discusses the use of D-optimal designs in the design of experiments (DOE) and artificial neural networks (ANN) in predicting the deflection and stresses of carbon fibre reinforced plastic (CFRP) square laminated composite plate subjected to uniformly distributed load. For training and testing of the ANN model, a number of finite element analyses have been carried out using D-optimal designs by varying the fibre orientations and thickness of each lamina. The composite plate is modeled using shell 99 elements. The ANN model has been developed using multilayer perceptron (MLP) backpropagation algorithm. The adequacy of the developed model is verified by root mean square error and regression coefficient. The results showed that the training algorithm of backpropagation was sufficient enough in predicting the deflection and stresses.