Backpropagation Technique Research Articles

On the Use of an Equivalent Currents-Based Technique to Improve Electromagnetic Imaging

Accurate imaging from scattered field measurements is one of the main research topics in the field of Non-Destructive Testing. Different methodologies for inverse scattering have been developed, ranging from backpropagation techniques to full-wave inversion methods. While far field approaches are valid for the majority of the imaging systems, there are some scenarios where these approaches do not provide accurate modeling of the field radiated by the antennas of the imaging system. This results in some kind of distortion in the resulting output microwave images. To address this issue, proper characterization and modeling of the fields radiated by the antennas is required. This contribution proposes the use of an equivalent currents-based technique to characterize the transmitting and receiving antennas of the imaging sensor. The goal is to calculate accurately the field radiated by these antennas within the imaging domain, using it in the backpropagation algorithm. The methodology described in this contribution is supported by measurements for monostatic and multistatic configurations.

An Improved Image Processing Based on Deep Learning Backpropagation Technique

In terms of image processing, encryption plays the main role in the field of image transmission. Using one algorithm of deep learning (DL), such as neural network backpropagation, increases the performance of encryption by learning the parameters and weights derived from the image itself. The use of more than one layer in the neural network improves the performance of the algorithm. Also, in the process of image encryption, randomness is an important component, especially when used by smart learning methods. Deep neural networks are related to pixels used to manipulate position and value according to the predicted new value given from a variable neural system. It also includes messy encrypted images used via applying randomness and increasing the key space in addition to using the logistic and Henon map for complexity. The main goal of any encryption method is to increase the complexity of the encrypted image to be difficult or impossible to decrypt the image without the proposed key. One of the important measurements for image encryption is the histogram and how it can be uniformed by the proposed method. Variables of randomness are used as features for the deep learning system, with feedback during iteration. An ideal image processing encryption yields high messy images by keeping the quality. Experimental results showed the backpropagation algorithm achieved better results than other algorithms.

Machine Learning Empowered Software Defect Prediction System

Production of high-quality software at lower cost has always been the main concern of developers. However, due to exponential increases in size and complexity, the development of qualitative software with lower costs is almost impossible. This issue can be resolved by identifying defects at the early stages of the development lifecycle. As a significant amount of resources are consumed in testing activities, if only those software modules are shortlisted for testing that is identified as defective, then the overall cost of development can be reduced with the assurance of high quality. An artificial neural network is considered as one of the extensively used machine-learning techniques for predicting defect-prone software modules. In this paper, a cloud-based framework for real-time software-defect prediction is presented. In the proposed framework, empirical analysis is performed to compare the performance of four training algorithms of the back-propagation technique on software-defect prediction: Bayesian regularization (BR), Scaled Conjugate Gradient, Broyden–Fletcher–Goldfarb–Shanno Quasi-Newton, and Levenberg-Marquardt algorithms. The proposed framework also includes a fuzzy layer to identify the best training function based on performance. Publicly available cleaned versions of NASA datasets are used in this study. Various measures are used for performance evaluation including specificity, precision, recall, F-measure, an area under the receiver operating characteristic curve, accuracy, R2, and mean-square error. Two graphical user interface tools are developed in MatLab software to implement the proposed framework. The first tool is developed for comparing training functions as well as for extracting the results; the second tool is developed for the selection of the best training function using fuzzy logic. A BR training algorithm is selected by the fuzzy layer as it outperformed the others in most of the performance measures. The accuracy of the BR training function is also compared with other widely used machine-learning techniques, from which it was found that the BR performed better among all training functions.

Cylindrical MIMO-SAR Imaging and Associated 3-D Fourier Processing

In this paper, we propose a new three dimensional cylindrical imaging system based on the combination of a multiple-input multiple-output arc array in azimuth and a synthetic aperture radar on the vertical dimension for applications focused on buried threat detection. Imaging algorithms are adapted to these specific transceiver distributions in order to accelerate reconstructions in the Fourier domain. Following theoretical developments, an experimental proof of concept is proposed at 6 GHz using a circular dual positioning system specifically designed for these new studies. The validity of the proposed techniques is extended to larger cylindrical arrays using numerical simulations of complex volumetric scenes in the 5–12 GHz band. This work allows for significant gains in computation time for these specific imaging systems, which are more than three orders of magnitude shorter than those of conventional backpropagation techniques in the studied cases composed of more than 250,000 voxels.

Developing an artificial neural network for detecting COVID-19 disease.

BACKGROUND:From December 2019, atypical pneumonia termed COVID-19 has been increasing exponentially across the world. It poses a great threat and challenge to world health and the economy. Medical specialists face uncertainty in making decisions based on their judgment for COVID-19. Thus, this study aimed to establish an intelligent model based on artificial neural networks (ANNs) for diagnosing COVID-19.MATERIALS AND METHODS:Using a single-center registry, we studied the records of 250 confirmed COVID-19 and 150 negative cases from February 9, 2020, to October 20, 2020. The correlation coefficient technique was used to determine the most significant variables of the ANN model. The variables at P < 0.05 were used for model construction. We applied the back-propagation technique for training a neural network on the dataset. After comparing different neural network configurations, the best configuration of ANN was acquired, then its strength has been evaluated.RESULTS:After the feature selection process, a total of 18 variables were determined as the most relevant predictors for developing the ANN models. The results indicated that two nested loops' architecture of 9-10-15-2 (10 and 15 neurons used in layer 1 and layer 2, respectively) with the area under the curve of 0.982, the sensitivity of 96.4%, specificity of 90.6%, and accuracy of 94% was introduced as the best configuration model for COVID-19 diagnosis.CONCLUSION:The proposed ANN-based clinical decision support system could be considered as a suitable computational technique for the frontline practitioner in early detection, effective intervention, and possibly a reduction of mortality in patients with COVID-19.

Artificial Intelligence in Geospatial Analysis for Flood Vulnerability Assessment: A Case of Dire Dawa Watershed, Awash Basin, Ethiopia.

This study presents the novelty artificial intelligence in geospatial analysis for flood vulnerability assessment in Dire Dawa, Ethiopia. Flood-causing factors such as rainfall, slope, LULC, elevation NDVI, TWI, SAVI, K-factor, R-factor, river distance, geomorphology, road distance, SPI, and population density were used to train the ANN model. The weights were generated in the ANN model and prioritized. Initial values were randomly assigned to the NN and trained with the feedforward processes. Ground-truthing points collected from the historical flood events of 2006 were used as targeting data during the training. A rough flood hazard map generated in feedforward was compared with the actual data, and the errors were propagated back into the NN with the backpropagation technique, and this step was repeated until a good agreement was made between the result of the GIS-ANN and the historical flood events. The results were overlapped with ground-truthing points at 88.46% and 89.15% agreement during training and validation periods. Therefore, the application of the GIS-ANN for the assessment of flood vulnerable zones for this city and its catchment was successful. The result of this study can also be further considered along with the city and its catchment for practical flood management.

Preparation of many-body ground states by time evolution with variational microscopic magnetic fields and incomplete interactions

State preparation is of fundamental importance in quantum physics, which can be realized by constructing the quantum circuit as a unitary that transforms the initial state to the target, or implementing a quantum control protocol to evolve to the target state with a designed Hamiltonian. In this work, we study the latter on quantum many-body systems by the time evolution with fixed couplings and variational magnetic fields. In specific, we consider to prepare the ground states of the Hamiltonians containing certain interactions that are missing in the Hamiltonians for the time evolution. An optimization method is proposed to optimize the magnetic fields by "fine-graining" the discretization of time, in order to gain high precision and stability. The back propagation technique is utilized to obtain the gradients of the fields against the logarithmic fidelity. Our method is tested on preparing the ground state of Heisenberg chain with the time evolution by the XY and Ising interactions, and its performance surpasses two baseline methods that use local and global optimization strategies, respectively. Our work can be applied and generalized to other quantum models such as those defined on higher dimensional lattices. It enlightens to reduce the complexity of the required interactions for implementing quantum control or other tasks in quantum information and computation by means of optimizing the magnetic fields.

Bayesian Convolutional Neural Networks for Seismic Facies Classification

The seismic response of geological reservoirs is a function of the elastic properties of porous rocks, which depends on rock types, petrophysical features, and geological environments. Such rock characteristics are generally classified into geological facies. We propose to use the convolutional neural networks in a Bayesian framework to predict facies based on seismic data and quantify the uncertainty in the classification. A variational approach is adopted to approximate the posterior distribution of neural parameters that is mathematically intractable. The network is trained on labeled examples. The mean and the standard deviation of the distribution of neural parameters are randomly drawn from predefined Gaussian functions for the initialization, and are updated by minimizing the negative evidence lower bound. The facies classification is applied to seismic sections not included in the training data set. We draw multiple random samples from the trained variational posterior distribution to simulate an ensemble predictor and classify the most probable seismic facies. We implement the proposed network in the open-source library of TensorFlow Probability, for its convenience and flexibility. The applications show that the internal regions of the seismic sections are generally classified with higher confidence than their boundaries, as measured by the predictive entropy that is calculated based on a multiclass probability across the possible facies. A plain neural network is also applied for comparison, by assigning fixed values to the neural parameters using a classical backpropagation technique. The comparison shows consistent results; however, the proposed approach is able to assess the uncertainty in the predictions.

Conditioning surface-based geological models to well data using artificial neural networks

Surface-based modelling provides a computationally efficient approach for generating geometrically realistic representations of heterogeneity in reservoir models. However, conditioning Surface-Based Geological Models (SBGMs) to well data can be challenging because it is an ill-posed inverse problem with spatially distributed parameters. To aid fast and efficient conditioning, we use here SBGMs that model geometries using parametric, grid-free surfaces that require few parameters to represent even realistic geological architectures. A neural network is trained to learn the underlying process of generating SBGMs by learning the relationship between the parametrized SBGM inputs and the resulting facies identified at well locations. To condition the SBGM to these observed data, inverse modelling of the SBGM inputs is achieved by replacing the forward model with the pre-trained neural network and optimizing the network inputs using the back-propagation technique applied in training the neural network. An analysis of the uncertainties associated with the conditioned realisations demonstrates the applicability of the approach for evaluating spatial variations in geological heterogeneity away from control data in reservoir modelling. This approach for generating geologically plausible models that are calibrated with observed well data could also be extended to other geological modelling techniques such as object- and process-based modelling.

MHD Boundary Layer Flow over a Stretching Sheet: A New Stochastic Method

In this study, a new computing model is developed using the strength of feed-forward neural networks with the Levenberg–Marquardt scheme-based backpropagation technique (NN-BLMS). It is used to find a solution for the nonlinear system obtained from the governing equations of the magnetohydrodyanmic (MHD) boundary layer flow over a stretching sheet. Moreover, the partial differential equations (PDEs) for the MHD boundary layer flow over a stretching sheet are converting into ordinary differential equations (ODEs) with the help of similarity transformation. A dataset for the proposed NN-BLMM-based model is generated at different scenarios by a variation of various embedding parameters: Deborah number β and magnetic parameter (M). The training (TR), testing (TS), and validation (VD) of the NN-BLMS model are evaluated in the generated scenarios to compare the obtained results with the reference results. For the fluidic system convergence analysis, a number of metrics, such as the mean square error (MSE), error histogram (EH), and regression (RG) plots, are utilized for measuring the effectiveness and performance of the NN-BLMS infrastructure model. The experiments showed that comparisons between the results of proposed model and the reference results match in terms of convergence up to E-02 to E-10. This proves the validity of the NN-BLMS model. Furthermore, the results demonstrated that there is a decrease in the thickness of the boundary layer by increasing the Deborah number and magnetic parameter. The importance of the experiment can be seen due to its industrial applications such as MHD power generation, MHD generators, and MHD pumps.

A New ANFIS-Based Hybrid Method in the Design and Fabrication of a High-Performance Novel Microstrip Diplexer for Wireless Applications

In this work, we have used a novel adaptive neuro-fuzzy inference system (ANFIS) method to design and fabricate a high-performance microstrip diplexer. For developing the proposed ANFIS model, the hybrid learning method consisting of least square estimation and back-propagation (BP) techniques is utilized. To achieve a compact diplexer, a designing process written in MATLAB 7.4 software is introduced based on the proposed ANFIS model. The basic microstrip resonator used in this study is mathematically analyzed. The designed microstrip diplexer operates at 2.2[Formula: see text]GHz and 5.1[Formula: see text]GHz for wideband wireless applications. Compared to the previous works, it has the minimum insertion losses and the smallest area of 0.007 [Formula: see text] (72.2[Formula: see text]mm2). It has flat channels with very low group delays (GDs) and wide fractional bandwidths (FBWs). The GDs at its lower and upper channels are only 0.48[Formula: see text]ns and 0.76[Formula: see text]ns, respectively. Another advantage of this work is its suppressed harmonics up to 12.9[Formula: see text]GHz (5th harmonic). To design the proposed diplexer, an LC model of the presented resonator is introduced and analyzed. To verify the simulation results and the presented ANFIS method, we fabricated and measured the proposed diplexer. The results show that both simulations and measurements data are in good agreement, which give reliability to the proposed ANFIS method.

A machine learning approach to estimate magnetorheological suspension composition based on magnetic field dependent-rheological properties

This paper presents an inverse model of magnetorheological (MR) suspensions to predict the compositions as a function of magnetic field-dependent rheological properties using the feedforward neural network (FFNN) model. Although many variations of MR suspension compositions have been published, the composition of the MR suspensions needs to be chosen manually by considering the required rheological properties. Therefore, this paper proposed a systematic method to predict MR suspension composition based on the rheological properties by employing extreme learning machine (ELM) and backpropagation (BP) techniques to build FFNN models. The model is built based on three experimental datasets representing particle shapes, sizes, and weight percentages. Three input topologies are proposed involving yield stress at the on-state condition, yield stress at the off-state condition , yield stress slope , and magnetic fields . The outputs are the particle weight percentages, milling time of magnetic particles representing the particle shapes, and particle sizes for each dataset model. FFNN models trained by ELM and BP have shown comparable accuracy. The simulations have also shown that the inclusion of the slope as one of the model inputs can produce R 2 of more than 0.80 for training and testing data. The off-state condition of yield stress can also improve the model performances if the off-state yield stress has a high correlation with the output. Finally, from the results, it can be concluded that the proposed machine learning approaches and topologies have successfully predicted the compositions in good agreement with the experimental data.

AN EXAMINATION OF EXPERT SYSTEM NEURAL NETWORK AND DATA MINING FOR CARDIAC DISEASES

Now a day’s neural network are being used successfully in an increasing number of application areas. Expert system is used comprehensively in several research domains. Heart disease diagnosis is a complicated process and requires high level of expertise. This paper examines and aiming to study a expert system for diagnosing of heart disease using support vector machine, feed forward Back propagation technique, Radial Basis Function, Generalized Regression Neural Network etc with other Data mining techniques that are used for diagnosis for heart diseases. This paper studies the analysis and examines the different approaches used for the detection of heart disease using different neural network techniques and data mining techniques.

Falkner–Skan Equation with Heat Transfer: A New Stochastic Numerical Approach

In this study, a new computing model is developed using the strength of feedforward neural networks with the Levenberg–Marquardt method- (NN-BLMM-) based backpropagation technique. It is used to find a solution for the nonlinear system obtained from the governing equations of Falkner–Skan with heat transfer (FSE-HT). Moreover, the partial differential equations (PDEs) for the unsteady squeezing flow of heat and mass transfer of the viscous fluid are converted into ordinary differential equations (ODEs) with the help of similarity transformation. A dataset for the proposed NN-BLMM-based model is generated in different scenarios by a variation of various embedding parameters, Deborah number ( β ) and Prandtl number (Pr). The training (TR), testing (TS), and validation (VD) of the NN-BLMM model are evaluated in the generated scenarios to compare the obtained results with the reference results. For the fluidic system convergence analysis, a number of metrics such as the mean square error (MSE), error histogram (EH), and regression (RG) plots are utilized for measuring the effectiveness and performance of the NN-BLMM infrastructure model. The experiments showed that comparisons between the results of the proposed model and the reference results match in terms of convergence up to E-05 to E-10. This proves the validity of the NN-BLMM model. Furthermore, the results demonstrated that there is an increase in the velocity profile and a decrease in the thickness of the thermal boundary layer by increasing the Deborah number. Also, the thickness of the thermal boundary layer is decreased by increasing the Prandtl number.

Dexterity analysis and intelligent trajectory planning of redundant dual arms for an upper body humanoid robot

PurposeMost of the redundant dual-arm robots are singular free, dexterous and collision free compared to other robotic arms. This paper aims to analyse the workspace of redundant arms to study the manipulability. Furthermore, multi-layer perceptron (MLP) algorithm is used to determine the various joint parameters of both the upper body redundant arms. Trajectory planning of robotic arms is carried out with the help of inverse solutions obtained from the MLP algorithm.Design/methodology/approachIn this paper, the kinematic equations are derived from screw theory approach and inverse kinematic solutions are determined using MLP algorithm. Levenberg–Marquardt (LM) and Bayesian regulation (BR) techniques are used as the backpropagation algorithms. The results from two backpropagation techniques are compared for determining the prediction accuracy. The inverse solutions obtained from the MLP algorithm are then used to optimize the cubic spline trajectories planned for avoiding collision between arms with the help of convex optimization technique. The dexterity of the redundant arms is analysed with the help of Cartesian workspace of arms.FindingsDexterity of redundant arms is analysed by studying the voids and singular spaces present inside the workspace of arms. MLP algorithms determine unique solutions with less computational effort using BR backpropagation. The inverse solutions obtained from MLP algorithm effectively optimize the cubic spline trajectory for the redundant dual arms using convex optimization technique.Originality/valueMost of the MLP algorithms used for determining the inverse solutions are used with LM backpropagation technique. In this paper, BR technique is used as the backpropagation technique. BR technique converges fast with less computational time than LM method. The inverse solutions of arm joints for traversing optimized cubic spline trajectory using convex optimization technique are computed from the MLP algorithm.

Kajian Adaptive Neuro-Fuzzy Inference System (ANFIS) Dalam Memprediksi Penerimaan Mahasiswa Baru Pada Universitas Buana Perjuangan Karawang

Abstract - The process of admitting new students is an annual routine activity that occurs in a university. This activity is the starting point of the process of searching for prospective new students who meet the criteria expected by the college. One of the colleges that holds new student admissions every year is Buana Perjuangan University, Karawang. There have been several studies that have been conducted on predictions of new students by other researchers, but the results have not been very satisfying, especially problems with the level of accuracy and error. Research on ANFIS studies to predict new students as a solution to the problem of accuracy. This study uses two ANFIS models, namely Backpropagation and Hybrid techniques. The application of the Adaptive Neuro-Fuzzy Inference System (ANFIS) model in the predictions of new students at Buana Perjuangan University, Karawang was successful. Based on the results of training, the Backpropagation technique has an error rate of 0.0394 and the Hybrid technique has an error rate of 0.0662. Based on the predictive accuracy value that has been done, the Backpropagation technique has an accuracy of 4.8 for the value of Mean Absolute Deviation (MAD) and 0.156364623 for the value of Mean Absolute Percentage Error (MAPE). Meanwhile, based on the Mean Absolute Deviation (MAD) value, the Backpropagation technique has a value of 0.5 and 0.09516671 for the Mean Absolute Percentage Error (MAPE) value. So it can be concluded that the Hybrid technique has a better level of accuracy than the Backpropation technique in predicting the number of new students at the University of Buana Perjuangan Karawang.&#x0D; &#x0D; Keywords: ANFIS, Backpropagation, Hybrid, Prediction

Performance Evaluation of Adaptive Neuro-Fuzzy Inference System (ANFIS) In Predicting New Students (Case Study : UBP Karawang)

The process of admitting new students is an annual routine activity that occurs in a university. This activity is the starting point of the process of searching for prospective new students who meet the criteria expected by the college. One of the colleges that holds new student admissions every year is Buana Perjuangan University, Karawang. There have been several studies that have been conducted on predictions of new students by other researchers, but the results have not been very satisfying, especially problems with the level of accuracy and error. Research on ANFIS studies to predict new students as a solution to the problem of accuracy. This study uses two ANFIS models, namely Backpropagation and Hybrid techniques. The application of the Adaptive Neuro-Fuzzy Inference System (ANFIS) model in the predictions of new students at Buana Perjuangan University, Karawang was successful. Based on the results of training, the Backpropagation technique has an error rate of 0.0394 and the Hybrid technique has an error rate of 0.0662. Based on the predictive accuracy value that has been done, the Backpropagation technique has an accuracy of 4.8 for the value of Mean Absolute Deviation (MAD) and 0.156364623 for the value of Mean Absolute Percentage Error (MAPE). Meanwhile, based on the Mean Absolute Deviation (MAD) value, the Backpropagation technique has a value of 0.5 and 0.09516671 for the Mean Absolute Percentage Error (MAPE) value. So it can be concluded that the Hybrid technique has a better level of accuracy than the Backpropation technique in predicting the number of new students at the University of Buana Perjuangan Karawang

The prediction of wellhead pressure for multiphase flow of vertical wells using artificial neural networks

Multiphase flow through both vertical and horizontal tubulars is getting higher interest in the oil and gas industry. Prediction of wellhead pressure through vertical wells is a very critical point that has a great influence on different applications. In this research, an artificial neural network with backpropagation technique (ANN-BP) was used to predict the wellhead pressure (WHP) for multiphase flow for vertical well systems. This permits the calculation of the pressure drop across the vertical well section by knowing the bottom hole flowing pressure (BHP). More than 150 data sets from different wells in the Middle East with different conditions were used to build the model. About 80% of the data were used to train the model while the rest unseen 20% were used to test and validate the model. The network structure, including the training function, the transfer function, the number of hidden layers, and the number of neurons in each layer, was highly optimized by trying different combinations of each parameter. The developed ANN model yielded high accuracy in predicting the WHP with an average absolute percentage error (AAPE) for both training and testing which are 0.61% and 1.13%, respectively. The optimized model comprised a single hidden layer with 20 neurons activated with the transfer function “tansig.” The correlation coefficient between the actual and predicted values for both training and testing was 0.98. A new empirical equation was then developed to mimic the developed ANN model by extracting the network weights and biases. The developed ANN-based correlation outweighs the previously established correlations in the literature upon comparison using unseen dataset.

Robust ANN-Based Control of Modified PUC-5 Inverter for Solar PV Applications

Conventional PI controllers are vulnerable to changes in parameters and are difficult to tune. In this work, an artificial neural network (ANN) based controller is developed for the robust operation of a single-phase modified packed U-cell five-level inverter (MPUC-5) for solar PV application under variable insolation conditions. An MPUC-5 is a converter with a main and an auxiliary dc link of equal magnitude; although five-level operation is also still feasible with different voltages also. The maximum power point (MPP) of a PV array changes with the variation in the solar insolation. This results in a variable voltage at the output of the boost converter while maintaining the load line at the MPP. Consequently, the fundamental value of the output of the MPUC-5 also tends to change. Thus, it is required to produce angles that commit to an ac output voltage with a constant fundamental value and constrained to a minimum total harmonic distortion along with a third-order harmonic mitigation as per the grid codes, irrespective of the change in the dc-link voltages. A genetic algorithm is employed for this purpose. A large dataset is prepared for two-angle and four-angle operation of MPUC-5 under various dc-link voltages and constraints with which an ANN-based controller is trained. A neural network with a hidden layer is trained with the backpropagation technique; and once a correlation is developed, the network can be operated for a wide range of operating conditions. The robustness of the controller is verified through simulation in MATLAB/Simulink environment and validated by experimental emulation in an hardware in loop environment.

Array-Based Guided Wave Source Location Using Dispersion Compensation

Abstract An important advantage of guided waves is their ability to propagate large distances and yield more information about flaws than bulk waves. Unfortunately, the multi-modal, dispersive nature of guided waves makes them difficult to use for locating flaws. In this work, we present a method and experimental data for removing the deleterious effects of multi-mode dispersion allowing for source localization at frequencies comparable to those of bulk waves. Time domain signals are obtained using a novel 64-element phased array and processed to extract wave number and frequency spectra. By an application of Auld’s electro-mechanical reciprocity relation, mode contributions are extracted approximately using a variational method. Once mode contributions have been obtained, the dispersion for each mode is removed via back-propagation techniques. Excepting the presence of a small artifact at high frequency-thicknesses, experimental data successfully demonstrate the robustness and viability of this approach to guided wave source location.