Levenberg-Marquardt Learning Algorithm Research Articles

A design of license plate recognition system using convolutional neural network

This paper proposes an improved Convolutional Neural Network (CNN) algorithm approach for license plate recognition system. The main contribution of this work is on the methodology to determine the best model for four-layered CNN architecture that has been used as the recognition method. This is achieved by validating the best parameters of the enhanced Stochastic Diagonal Levenberg Marquardt (SDLM) learning algorithm and network size of CNN. Several preprocessing algorithms such as Sobel operator edge detection, morphological operation and connected component analysis have been used to localize the license plate, isolate and segment the characters respectively before feeding the input to CNN. It is found that the proposed model is superior when subjected to multi-scaling and variations of input patterns. As a result, the license plate preprocessing stage achieved 74.7% accuracy and CNN recognition stage achieved 94.6% accuracy.

Controlling Electronic Devices with Brain Rhythms/Electrical Activity Using Artificial Neural Network (ANN).

The purpose of this research study was to explore the possibility to develop a brain-computer interface (BCI). The main objective was that the BCI should be able to recognize brain activity. BCI is an emerging technology which focuses on communication between software and hardware and permitting the use of brain activity to control electronic devices, such as wheelchairs, computers and robots. The interface was developed, and consists of EEG Bitronics, Arduino and a computer; moreover, two versions of the BCIANNET software were developed to be used with this hardware. This BCI used artificial neural network (ANN) as a main processing method, with the Butterworth filter used as the data pre-processing algorithm for ANN. Twelve subjects were measured to collect the datasets. Tasks were given to subjects to stimulate brain activity. The purpose of the experiments was to test and confirm the performance of the developed software. The aim of the software was to separate important rhythms such as alpha, beta, gamma and delta from other EEG signals. As a result, this study showed that the Levenberg–Marquardt algorithm is the best compared with the backpropagation, resilient backpropagation, and error correction algorithms. The final developed version of the software is an effective tool for research in the field of BCI. The study showed that using the Levenberg–Marquardt learning algorithm gave an accuracy of prediction around 60% on the testing dataset.

Modeling callus induction and regeneration in an anther culture of tomato (Lycopersicon esculentum L.) using image processing and artificial neural network method.

Doubled haploids, subsequent to haploid induction, have wide range of applications in basic and applied plant studies. Various parameters can affect the efficiency of haploid induction through an anther culture of tomato. The hybrid system of image processing-artificial neural network (ANN) was used to better understand callus induction and regeneration in an anther culture of tomato. The effect of parameters such as plant genotype, the concentrations of 2,4-dichlorophenoxyacetic acid (2,4-D) and kinetin (Kin) plant growth regulators, the concentration of gum arabic (GA) additive, the cold pretreatment duration, and flower length on callus induction percentage and number of regenerated calli in an anther culture of tomato were studied using multiple linear regression (MLR) and ANN models. The precise flower bud length was measured using an image processing technique. The 4',6-diamidino-2-phenylindole (DAPI) analysis showed that the flowers with 5-6.9mm length had the highest percentage of the mid- to late-uninucleate microspore stage. The best ANN model for both callus induction percentage and number of regenerated calli was a model with one hidden layer, 12-15 neurons in the first hidden layer, Levenberg-Marquardt learning algorithm, and Tan-Sigmoid transfer function in hidden layer, based on the root mean square error (RMSE), mean absolute error (MAE), and coefficient of determination (R2) statistics. The scatter plot of measured values versus the predicted values showed the superiority of the ANN to MLR model to predict the callus induction percentage in an anther culture of tomato. The sensitivity analysis of MLR and ANN models revealed the plant genotype and 2,4-D concentration as the most important factors affecting both callus induction percentage and number of regenerated calli. Since tomato is a recalcitrant plant to androgenesis-based pathway of haploid induction, therefore the results of the present study can be helpful to develop an efficient haploid induction protocol in tomato through an anther culture pathway.

Comparison of linear regression and artificial neural network technique for prediction of a soybean biodiesel yield

ABSTRACTAtmospheric pollution is one of the biggest problems all over the world. For this reason, researchers try to find alternative fuels for diesel engine, and biodiesel is the most feasible alternate fuel for diesel engines. In this study, linear regression (LR) and artificial neural network (ANN) used to predict the biodiesel yield produced by transesterification of soybean oil at constant temperature is reported in present work describes. The ANN estimation was done using a Levenberg–Marquardt learning algorithm (trainlm) with log sigmoid (logsig) neural network algorithm with 4 neurons in the hidden layer (3:4:1 topology). The experimental and ANN values were compared for the biodiesel yield. The value R2 = 0.9899 for ANN and R2 = 0.4198 for LR. Root mean square errors (RMSE) for ANN and LR are 0.6331 and 3.052, respectively. Results were compared with LR modeling. As a result, ANN gave more accurate results than LR and can be suggested as good a prediction method. It was followed by Fourier transform infrared (FTIR) spectroscopy analysis.

Predicting students\u2019 final degree classification using an extended profile

The students’ progression and attainment gap are considered as key performance indicators of many universities worldwide. Therefore, universities invest significantly in resources to reduce the attainment gap between good and poor performing students. In this regard, various mathematical models have been utilised to predict students’ performances in the hope of informing the support team to intervene at an early stage of the at risk student’s at the university. In this work, we used a combination of institutional, academic, demographic, psychological and economic factors to predict students’ performances using a multi-layered neural network (NN) to classify students’ degrees into either a good or basic degree class. To our knowledge, the usage of such an extended profile is novel. A feed-forward network with 100 nodes in the hidden layer trained using Levenberg-Marquardt learning algorithm was able to achieve the best performance with an average classification accuracy of 83.7%, sensitivity of 77.37%, specificity of 85.16%, Positive Predictive Value of 94.04%, and Negative Predictive Value of 50.93%. The NN model was also compared against other classifiers specifically k-Nearest Neighbour, Decision Tree and Support Vector Machine on the same dataset using the same features. The results indicate that the NN outperforms all other classifiers in terms of overall classification accuracy and shows promise for the method to be used in Student Success ventures in the universities in an automatic manner.

RETRACTED ARTICLE: Fuzzy rank correlation-based segmentation method and deep neural network for bone cancer identification

Bone malignant tumors are one of the important health problems because tumors are formed due to the affectedness of the healthiest bone tissues. This serious bone cancer has been identified with the help of the different risk factors such as chills, swear symptoms, swelling, bones weaken risk, and night swears symptoms. These symptoms are not easy to predict in beginning stage with accurate manner. So, automatic bone cancer detection system has been developed to predict the cancer in earlier sate. Initially, the bone images are collected from patient, and noise in the images is eliminated using median filter. After eliminating the noise, affected tumor part is detected by applying the intuitionistic fuzzy rank correlation. From the detected intuitionistic fuzzy-based clustered images, different statistical features are extracted. The derived features are processed by applying the deep neural networks layers that successfully examines each features using Levenberg–Marquardt learning algorithm. The successful learning process predicts bone cancer-related features with accurate manner (99.1%). Finally, excellence of bone cancer prediction system is analyzed using MATLAB-based experimental setup and performance metrics such as F1 score, accuracy, error rate and so on.

Short-term electric load forecasting in Tunisia using artificial neural networks

The accuracy of short-term electricity load forecasting is of great interest since it allows avoiding unexpected blackouts and lowering operating costs. In this paper, we aim to implement the artificial neural networks to model and forecast the half-hourly electric load demand in Tunisia over the period 2000–2008. To improve the quality of forecasts, the proposed artificial neural network model uses not only past electric load values as inputs, but also climatic and calendar variables. To determine the optimal structure of the neural network model, this paper employs the pattern search algorithm. Moreover, the neural network model is equipped with the Levenberg–Marquardt learning algorithm. Our findings confirm the performance of this algorithm to the view of evaluation indicators since the mean absolute percentage error values range between 1.1 and 3.4%. The analysis also shows the superiority of the Levenberg–Marquardt algorithm compared to the resilient back propagation algorithm and the conjugate gradient algorithm. In the light of the current research, we stress the aptness of the proposed artificial neural network model in forecasting short-term electricity demand.

Determination of Neural Network Parameters for Path Loss Prediction in Very High Frequency Wireless Channel

It is very important to understand the input features and the neural network parameters required for optimal path loss prediction in wireless communication channels. In this paper, an extensive investigation was conducted to determine the most appropriate neural network parameters for path loss prediction in Very High Frequency (VHF) band. Field measurements were conducted in an urban propagation environment to obtain relevant geographical and network information about the receiving mobile equipment and quantify the path losses of radio signals transmitted at 189.25 MHz and 479.25 MHz. Different neural network architectures were trained with varying kinds of input parameters, number of hidden neurons, activation functions, and learning algorithms to accurately predict corresponding path loss values. At the end of the experimentations, the performance of the developed Artificial Neural Network (ANN) models are evaluated using the following statistical metrics: Mean Absolute Error (MAE), Mean Squared Error (MSE), Root Mean Squared Error (RMSE), Standard Deviation (SD) and Regression coefficient (R). Results obtained show that the ANN model that yielded the best performance employed four input variables (latitude, longitude, elevation, and distance), nine hidden neurons, hyperbolic tangent sigmoid (tansig) activation function, and the Levenberg-Marquardt (LM) learning algorithm with MAE, MSE, RMSE, SD and R values of 0.58 dB, 0.66 dB, 0.81 dB, 0.56 dB and 0.99 respectively. Finally, a comparative analysis of the developed model with Hata, COST 231, ECC-33 and Egli models showed that ANN-based path loss model has better prediction accuracy and generalization ability than the empirical models.

Artificial neural networks and multiple linear regression as potential methods for modeling seed yield of safflower (Carthamus tinctorius L.)

Modeling relationships between safflower seed yield (SY) and its components would be helpful to understand the most important trait-(s) with significant effects on seed yield, which in turn helps safflower breeders in the indirect selection of high-performance cultivars. For this purpose, we evaluated performance of five different artificial neural network (ANN) models, including Generalized feed forward (GFF), multilayer perceptron (MLP), Jordan/Elman (JE), Principal component analysis (PCA), and Radial basis function (RBF) with different learning algorithms, transfer functions, hidden layers and neuron in each layer, along with multi-linear regression model to predict seed yield of safflower. A panel of 25 lines/cultivars was used during two years (2012–2014) to evaluate some of the important traits related to seed yield. PCA was used to perform feature extraction in the first stage of modeling. Based on PCA and first two components that incorporated approximately 79% of the total variability, five agro morphological and phenological traits, including plant height (PH), number of branches per plant (NBP), number of capsules per plant (NCP), thousand seed weight (TSW), and number of seeds per capsule (NSC), due to high values of Eigen vectors were selected as input. The performance of the models in predicting seed yield was determined using some of the statistical quality parameters, including coefficient of determination (R2), root mean square error (RMSE) and mean absolute error (MAE). The MLP model with the Levenberg–Marquardt learning algorithm, sigmoid axon transfer function, two hidden layers (i.e. topology 5-5-4-1) and 1500 training epochs was selected as the best model to predict seed yield of safflower. Sensitivity analysis indicated that the NCP is the most influential trait to predict SY in the MLP/ANN model.

Development of Feed-Forward Back-Propagation Neural Model to Predict the Energy and Exergy Analysis of Solar Air Heater

In the present work, Artificial Neural Network (ANN) model has been developed to predict the energy and exergy efficiency of a roughened solar air heater (SAH). Total fifty data sets of samples, obtained by conducting experiments on SAHs with three different specification of wire-rib roughness on the absorber plates, have been used in this work. These experimental data and calculated values of thermal efficiency and exergy efficiency have been used to develop an ANN model. Levenberg-Marquardt (LM) and Scaled Conjugate Gradient (SCG) learning algorithm were used to train the proposed ANN model. Six numbers of neurons were found with LM learning algorithm in the hidden layer as the optimal value on the basis of statistical error analysis. In the input layer, the time of experiments, mass flow rate, ambient temperature, mean temperature of air, absorber plate temperature and solar radiation intensity have been taken as input parameters; and energy efficiency and exergy efficiency have been taken as output parameters in the output layer. The 6-6-2 neural model has been obtained as the optimal model for prediction. Performance predictions using ANN were compared with the experimental data and a close agreement was observed. Statistical error analysis was used to evaluate the results. Citation:  Ghritlahre, H. K. (2018). Development of feed-forward back-propagation neural model to predict the energy and exergy analysis of solar air heater. Trends in Renewable Energy, 4, 213-235. DOI: 10.17737/tre.2018.4.2.0078

Closed Loop Temperature Control System Design Using Conventional and Advanced Intelligent Controller

Requirement of energy demand growing rapidly around the world due to modern suburban life styles and tendency for indoor comfort. These energy stipulation leads to design of intelligent control strategies to make the system energy efficient. Design an efficient room heater to make it energy proficient using intelligent control is an attractive research topic today. This article proposed a design of temperature controller with the purpose to sense the temperature of a room and maintain it to a fixed desired value. Conventional Proportional Integral (PI) controller as well as modern Artificial Neural Network (ANN) based PI controller have been used as the control strategies when a model of PI controller based on ANN with Levenberg-Marquardt learning algorithm has been introduced in this article. The simulation report of projected design has been shown and it has been concluded that the proposed control strategy is very advantageous due to its small overshoot, fast response, and high precision. All designs and simulations have been done in Simulink of Matllab platform.

The development and evaluation of a portable polyethylene biogas reactor

Several factors can influence the process of biogas production. The type of reactor is one of the key factors that influence biogas production. Therefore, the aim of this study was to construct a portable horizontal polyethylene-based biogas reactor. In addition, the performance of the developed biogas reactor was tested through digestion of cow manure. The experiments were carried out in Mashhad, Iran, during June–July 2016. Biogas production was studied over a span of 58 days’ hydraulic retention time. Artificial neural network (ANN) models were used to predict the production of biogas based on temperature and pH. The Levenberg–Marquardt learning algorithm was employed to develop the best model. The obtained biogas productivity was 0.27 m3 kgVS-1, indicating that the developed biogas reactor was optimum to convert the substrate into biogas. The ANN results highlighted that the best developed model consisted of an input layer with two input variables, one hidden layer with 15 neurons, and one output layer with the correlation coefficient of 0.90. Overall, it was concluded that the ANN models can be employed to prognosticate biogas production using a portable polyethylene biogas reactor.

Investigation of Performance of Tropospheric Ozone Estimations in The Industrial Region Using Differential Artificial Neural Networks Methods

The method of Levenberg-Marquardt learning algorithm was investigated for estimating tropospheric ozone concentration. The Levenberg-Marquardt learning algorithm has 12 input neurons (6 pollutants and 6 meteorological variables), 28 neurons in the hidden layer, and 1 output neuron for the Ozone (O3) estimate. The Multilayer Perceptron Model (MLP) performance was found to make good predictions with the mean square error (MSE) less than 1 µg/m3 (0.002 µg/m3). In addition, the correlation coefficient ranges from 0.74 to 0.95 in The Levenberg-Marquardt learning. The Levenberg-Marquardt learning algorithm that a multilayer perception method of Artificial Neural Network (ANN) has performed well and an effective approach for predicting tropospheric ozone. Ozone concentration was influenced predominantly by the nitrogen oxide (NOx, NO2, NO), SO2 and temperature. The model did not predict solar radiation to ozone with sufficient accuracy.

Horizontal Extrapolation of Wind Speed Distribution Using Neural Network for Wind Resource Assessment

To evaluate the wind potential on a site for future wind energy project, an accurate representation of the wind speed distribution is required. However, due to the lack of observations, wind engineers are conducted to use some statistical tools to estimate the characteristics of wind by the measurements from a nearby reference or data obtained from a short period. In this work, we aim at applying an information processing paradigm that is inspired by biological neurons, formal neurons, for the assessment of wind speed distribution. Two different learning algorithms are used so as to generate Artificial Neural Network with one hidden layer. Results prove that learning by means of Bayesian regularization, in comparison with Levenberg–Marquardt learning algorithm, gives the best performance. In addition, the proposed network allows significant results in horizontal wind extrapolation.

Effective prediction model for Hungarian small‐scale solar power output

Owing to critical role of photovoltaic (PV) power in oncoming energy market, an accurate PV power forecasting model is demanded. In this paper, an effective solar power prediction model composed of variational mode decomposition, information-theoretic feature selection, and forecasting engine with high learning capability is proposed. The feature selection method is based on information-theoretic criteria and an optimisation algorithm. The forecasting engine is multilayer perceptron neural network equipped with modified Levenberg–Marquardt learning algorithm. An evolutionary algorithm is also incorporated into the training mechanism of the forecasting engine to enhance its learning capability. Effectiveness of the proposed PV prediction model is illustrated on a Hungarian solar power plant.

Optimal configuration of multilayer perceptron neural network classifier for recognition of intracranial epileptic seizures

Abstract Background This paper presents optimized configuration of multilayer perceptron neural network (MLP-NN) as a pattern classifier for recognition of intracranial electroencephalogram (iEEG) epileptic seizures. Though qualitative analysis of intracranial recordings from the patients involves cumbersome procedure, it provides significant neuronal activities of the brain that is essential for clinical decision making. Methods This study considered three epileptic seizures stages, namely, pre-ictal (set C), post-ictal (set D) and epileptic (set E) iEEG from the University of Bonn, Germany database. Four entropies, Shannon, log energy, spectral and Renyi entropies were considered as features to evaluate MLP-NN network. Four classification tasks with the dataset were formed, namely CE, DE, CDE, and CD. In order to identify the optimal configuration of MLP-NN for classification, network parameters such as input-hidden layer activation/transfer functions, hidden-output layer activation/transfer functions, network training/learning functions, the number of hidden neurons were considered and the optimality was arrived based on the mean square error (MSE), classification accuracy (CA). The efficiency of the entropy features was exploited by the parameters, mutual range of coefficient ( γ ), p-value, and z-score, which showed the band discrimination for various classification tasks. Results Simulations results showed that the tan sigmoid, pure linear were found to be optimal input-hidden, hidden-output transfer function and Levenberg-Marquardt learning algorithm as optimal training function for all the four classification tasks. It was inferred from the proposed study that the CA indirectly varies with γ value, p value, and MSE, and directly varies with z-score. From the experimental study, the best CA of 97.68%, 94.56%, 84.58%, and 57.8% was obtained for case CE, DE, CDE, and CD respectively. It can be concluded that proposed features with optimally configured MLP-NN found to be helpful for real-time iEEG classification.

Experimental evaluation of artificial neural network for predicting drainage water and groundwater salinity at various drain depths and spacing

Drainage design parameters of drain depth and spacing are the pivotal factors affect the drain water quality by radial flow of underground water. In this study, artificial neural network modeling has been employed with Levenberg–Marquardt learning algorithm followed by Sigmoid Axon transfer function to anticipate the temporal changes in shallow groundwater and drain water salinities at various depths and spaces of drain installation. Calibration and validation of the model results were carried out based on data obtained from experimental model with 1.8 m long, 1 m wide, and 1.2 m high. In the model, drains were installed at 20, 40, and 60 cm depths and 60, 90, and 180 cm spaces. The values of error indices of RMSE and SE as well as R2 between measured and simulated shallow groundwater salinities were 5.27 dS/m, 0.12, and 0.96, respectively. These indexes for drain water salinity were obtained 0.72 dS/m, 0.096, and 0.99, respectively. The key results revealed that artificial neural network methodology has a reasonable accuracy on simulating temporal shallow groundwater and drain water salinities in different drain depth and drain spacing.

Artificial Neural Networks Modeling of Dynamic Adsorption From Aqueous Solution

The aim of this work is to use multilayered perceptron artificial neural networks (MLP-ANN) and multiple linear regressions (MLR) models to predict the dynamic adsorption of the complex system of adsorbent-adsorbate in solid-liquid phase. A set of 1859 data points were used. For the (MLP-ANN), nine neurons were used in the input layer, sixteen neurons at hidden layer and one was used in the output layer. Levenberg Marquardt learning (LM) algorithm, logarithmic sigmoid transfer function and linear transfer function were used at the hidden and output layer respectively. The comparison of the obtained results in term of root mean square error (RMSE) and correlation coefficient (R) using the (MLP-ANN) and MLR models revealed the superiority of the (MLP-ANN) model in predicting of dynamic adsorption process. The statistic results showed a correlation coefficient R = 0.991 with root mean square error RMSE= 0.0521for the (MLP-ANN) model and R= 0.80 with RMSE=0.237for the MLR model. Thus, it can be suggested that the artificial neural network model gave far better and more significant results.

A comparative study of three different learning algorithms applied to ANFIS for predicting daily suspended sediment concentration

The modeling and prediction of suspended sediment in a river are key elements in global water recourses and environment policy and management. In the present study, an Adaptive Neuro-Fuzzy Inference System model trained with the Levenberg-Marquardt learning algorithm is considered for time series modeling of suspended sediment concentration in a river. The model is trained and validated using daily river discharge and suspended sediment concentration data from the Schuylkill River in the United States. The results of the proposed method are evaluated and compared with similar networks trained with the common Hybrid and Back-Propagation algorithms, which are widely used in the literature for prediction of suspended sediment concentration. Obtained results demonstrate that models trained with the Hybrid and Levenberg-Marquardt algorithms are comparable in terms of prediction accuracy. However, the networks trained with the Levenberg-Marquardt algorithm perform better than those trained with the Hybrid approach.

Image correction for cone-beam computed tomography simulator using neural network corrector

In this article, a neural network corrector is proposed to correct the image shift, yielding the degradation of three-dimensional image reconstruction, for each slice captured by cone-beam computed tomography simulator. There are 3 degrees of freedom in tube module of simulator; the central point of tube module should be aligned with the central point of detector module to guarantee the accurate image projection. However, the mechanism manufacturing and assembling tolerance will let the above aim cannot be met. Here, a standard kit is made to measure the image shift by 1° step from −10° to 10°. The measure data will be the input training data of proposed neural network corrector, and the corrected translation position will be the output of neural network corrector. The Levenberg–Marquardt learning algorithm adjusts the connected weights and biases of the neural network using a supervised gradient descent method, such that the defined error function can be minimized. To avoid the problem of overfitting and improve the generalized ability of the neural network, Bayesian regularization is added to the Levenberg–Marquardt learning algorithm. After the training of neural network corrector, the different target position commands are fed into the neural network corrector. Then, the corrected data from neural network corrector are fed to be the new position command to verify the image correction performance. Moreover, a phantom kit is made to check the corrected performance of the neural network corrector. Finally, the experimental results verify that the image shift can be reduced by the neural network corrector.