Feed Forward Propagation Neural Network Research Articles

ICA-LightGBM Algorithm for Predicting Compressive Strength of Geo-Polymer Concrete

The main goal of the present study is to investigate the capability of hybridizing the imperialist competitive algorithm (ICA) with an intelligent, robust, and data-driven technique named the light gradient boosting machine (LightGBM) to estimate the compressive strength of geo-polymer concrete (CSGCo). The hyper-parameters of the LightGBM algorithm have been optimized based on ICA and its accuracy improved. The obtained results from the proposed hybrid ICA-LightGBM are compared with the traditional LightGBM model as well as four different topologies of artificial neural networks (ANN) comprising a multi-layer perceptron neural network (MLP), radial basis function (RBF), generalized feed-forward neural network (GFFNN), and Bayesian regularized neural network (BRNN). The results of these models were compared based on three evaluation indices of R2, RMSE, and VAF for providing an objective evaluation of the performance and capability of the predictive models. Concerning the outcomes, the ICA-LightGBM with the R2 of (0.9871 and 0.9805), RMSE of (0.4703 and 1.3137), and VAF of (98.5773 and 98.0397) for training and testing phases, respectively, was a superior predictor to estimate the CSGCo compared to the LightGBM with the R2 of (0.9488 and 0.9478), RMSE of (0.9532 and 2.1631), and VAF of (94.3613 and 94.5173); the MLP with the R2 of (0.9067 and 0.8959), RMSE of (1.3093 and 3.3648), and VAF of (88.9888 and 84.9125); the RBF with the R2 of (0.8694 and 0.8055), RMSE of (1.4703 and 5.0309), and VAF of (86.3122 and 66.1888); the BRNN with the R2 of (0.9212 and 0.9107), RMSE of (1.1510 and 2.6569), and VAF of (91.4168 and 90.5854); and the GFFNN with the R2 of (0.9144 and 0.8925), RMSE of (1.1525 and 2.9415), and VAF of (91.4092 and 88.9088). Hence, the proposed ICA-LightGBM algorithm can be efficiently used in anticipating the CSGCo.

Application of reliability-based back-propagation causality-weighted neural networks to estimate air-overpressure due to mine blasting

In the present study, the air-overpressure (AOp) due to mine blasting is predicted using an uncertainty intelligence method based on the Z-number reliability and fuzzy cognitive map (FCM). Hence, the cause-and-effect relationship between effective parameters on AOp was first determined by a team of experts. But, expert opinions are accompanied by uncertainty. Therefore, Z-number concept is used to solve the uncertainty of expert opinions and improve views. Notably, the relationships dependent on expert opinions. To overcome this problem, two learning algorithms called nonlinear Hebbian (NLH) and differential evolution (DE) algorithms integrated to decrease the dependence on experts’ views. An FCM based on NLH-DE and reliability information was first designed; then inputs’ weights were extracted during several simulations in training process. The reliability inputs’ weights were imported into back-propagation causality-weighted neural networks (BPCWNNs). The coefficient of determination (R2), root mean square error (RMSE), variance account for (VAF), and accuracy indices were employed as evaluations criteria to select the optimal topology of artificial neural networks (ANNs). Back propagation neural network (BPNN) and generalized feed forward neural network (GFFNN) models were also used to compare and analyze the performance of the developed BPCWNNs model. The results indicated that Z-number considerably improved the accuracy of the ANN model. Notably, the model assessment revealed that the proposed BPCWNNs model was accurate compared to the BPNN and GFFNN and selected as the most superior model. The BPCWNN is the model with high performance which can be used as an artificial intelligence technique for predicting blast-induced AOp.

A visualized hybrid intelligent model to delineate Swedish fine-grained soil layers using clay sensitivity

In the current paper, a hybrid model was developed to generate 3D delineated soil horizons using clay sensitivity (St) with 1 m depth intervals in a landslide prone area in the southwest of Sweden. A hybridizing process was carried out using generalized feed forward neural network (GFFN) incorporated with genetic algorithm (GA). The model was conducted by means of seven variables consisting of the geographical coordinates and piezocone penetration test data (CPTu). The output of model (St) as a description of the effect of soil disturbance on shear strength plays a significant role in landslides in Sweden and thus can be applied for site-specific evaluation. Therefore, the use of St-based models to delineate soil layers can be a cost-effective solution to improve geo-engineering design practices and assist in the reduction of related environmental risks, such as catastrophic landslide events or excavation failures. Evaluated model performance based on different applied soil classifications showed 4.38% improvement in the predictability level of GFFN-GA compared to optimum GFFN. Accordingly, delineated soil layers were evaluated using different criteria including previous landslides as well as supplementary geophysical and geotechnical investigations. The results show that the adopted hybrid GFFN-GA is an efficient tool that can potentially be applied to delineate soil horizons for the prediction of future events.

Segmentation of Cervical Cancer by OLHT Based DT-CWT Techniques

Every year, cervical cancer (CC) is the leading cause of death in women around the world. If detected early enough, this cancer can be treated, and patients will receive adequate care. This study introduces a novel ultrasound-based method for detecting CC. The Oriented Local Histogram Technique (OLHT) is used to improve the image corners in the cervical image (CI), and the Dual-Tree Complex Wavelet Transform (DT-CWT) is used to build a multi-resolution image (CI). Wavelet, and Local Binary Pattern are among the elements retrieved from this improved multi-resolution CI (LBP). The retrieved appearance is trained and tested using a feed-forward propagation neural network, and the ANFIS classifier is utilized to classify them. The purpose of this classifier is to distinguish between normal and pathological cervical pictures. Sensitivity is 97.52 percent, specificity is 99.46 percent, accuracy is 98.39 percent, precision is 97.48 percent, PPV is 97.38 percent, NPV is 92.27 percent, LRP is 141.81 percent, 0.0946 percent LRN, FPR is 96.82 percent, and NPR is 91.46 percent for the CC detection categorization. The proposed methodology outperforms standard CC identification and classification methodologies.

Non-Smooth Regularization: Improvement to Learning Framework Through Extrapolation

Deep learning architectures employ various regularization terms to handle different types of priors. Non-smooth regularization terms have shown promising performance in the deep learning architectures and a learning framework has recently been proposed to train autoencoders with such regularization terms. While this framework efficiently manages the non-smooth term during training through proximal operators, it is limited to autoencoders and suffers from low convergence speed due to several optimization sub-problems that must be solved in a row. In this paper, we address these issues by extending the framework to general feed-forward neural networks and introducing variable extrapolation which can dramatically increase the convergence speed in each sub-problem. We show that the proposed update rules converge to a critical point of the objective function under mild conditions. To compare the resulting framework with the previously proposed one, we consider the problem of training sparse autoencoders and robustifying deep neural architectures against both targeted and untargeted attacks. Simulations show superior performance in both convergence speed and final objective function value.

Early Breast Cancer Detection Utilizing Artificial Neural Network

Breast cancer is one of the most dangerous cancer diseases for women in worldwide. A Computeraided diagnosis system is very helpful for radiologist for diagnosing micro calcification patterns earlier and faster than typical screening techniques. Maximum breast cancer cells are eventually form a lump or mass called a tumor. Moreover, some tumors are cancerous and some are not cancerous. The cancerous tumors are called malignant and non-cancerous tumors are called benign. The benign tumors are not dangerous to health. But the unchecked malignant tumors have the ability to spread in other organs of the body. For that early detection of benign and malignant tumor is important for confining the death of breast cancer. In these research study different neural networks such as, Multilayer Perceptron (MLP) Neural Network, Jordan/Elman Neural Network, Modular Neural Network (MNN), Generalized Feed-Forward Neural Network (GFFNN), Self-Organizing Feature Map (SOFM) Neural Network, Support Vector Machine (SVM) Neural Network, Probabilistic Neural Network (PNN) and Recurrent Neural Network (RNN) are used for classifying breast cancer tumor. And compare the results of these networks to find the best neural network for detecting breast cancer. The networks are tested on Wisconsin breast cancer (WBC) database. Finally, the comparing result showed that Probabilistic Neural Network shows the best detection result than other networks.

A hybridized intelligence model to improve the predictability level of strength index parameters of rocks

In the current paper, the uniaxial compressive strength (UCS) and Young modulus (E) of rocks were predicted using a hybridized intelligence method. The model was developed using an optimum multi-objective generalized feedforward neural network (GFFN) incorporated with an imperialist competitive metaheuristic algorithm (ICA) and managed using 208 datasets of different physical and mechanical quarries from almost all over of Iran. Rock class, density, porosity, P-wave velocity, point load index and water absorption were datacenter components. The predictability and accuracy performance of the hybrid ICA-GFFN model were discussed using different error criteria and confusion matrixes. The observed 5.4% and at least 32% improvement in hybrid ICA-GFFN than GFFN and multivariate regression (MVR) demonstrated feasible and accurate enough tools that can effectively be applied for multi-objective prediction purposes. The influence of inputs on predicted outputs was also identified using two different sensitivity analyses.

Landslide susceptibility mapping using hybridized block modular intelligence model

Landslide susceptibility map (LSM) provide useful tool for decision makers in hazard mitigation concerns. In the present paper, a novel hybrid block-based neural network model (HBNN) for the purpose of producing high-resolution LSM was developed. This hybrid approach was found through the mixture of expert modular structures and divide-and-conquer strategy incorporated with genetic algorithm (GA). The introduced HBNN then was applied on southern part of Guilan province (north of Iran) using 14 causative factors covering topographic and geomorphologic features, and geological and tectonical factors as well as hydrology, land data, and climate conditions. The landslide inventory map was provided using a synergy work from monitored events, interpretation of aerial photographs, and carried out geotechnical investigations in the area as well as field surveys. To insight, the predictability of proposed HBNN was compared with two developed models using multilayer perceptrons (MLPs) and generalized feed forward neural network (GFFN). The generated LSM was validated using receiver operating characteristic (ROC) curves, statistical error indices, and sensitivity and weight analyses as well as monitored landslides. Based on the compared metrics, HBNN with 86.52% and 90.15% in prediction and success rate as well as 89.36% for precision-recall curve demonstrated more consistent tool for future landslide susceptibility zonation. This implies on capability of developed HBNN in producing higher resolution and more reliable LSM for urban and land-use planners.

Cascade Artificial Neural Networks for Modeling Economic Performance: A New Perspective

This paper discusses a new representation for efficiency frontier method through a proposed algorithm for augmented feed forward back propagation neural network models, to estimate the economic performance, and the effectiveness of macroeconomic policies in Egyptian economy, using a quarter time series data from 1990Q1 to 2019Q2. In this study I develop artificial neural network models - ANN - in line with the conditions of the Egyptian economy, by building an optimal efficiency frontier and then comparing the actual performance of the Egyptian economy with that limit, which includes the lowest possible variations for both inflation and output. As for the new contribution to this study, it is to calculate the optimal inflation rate and the optimal output level in the Egyptian economy through a model that combines the higher predictive power of feed forward neural network models, and the high explanatory power of a stationary or random walk stochastic models, in order to obtain the fitted values of the optimal output level, and the optimal inflation rate. It is clear from the results of the study, the extent of the essential congruence between the actual Egyptian economic performance during the study period and the economic performance index that was built through the new contribution of this study.

Artificial intelligence models to generate visualized bedrock level: a case study in Sweden

Assessment of the spatial distribution of bedrock level (BL) as the lower boundary of soil layers is associated with many uncertainties. Increasing our knowledge about the spatial variability of BL through high resolution and more accurate predictive models is an important challenge for the design of safe and economical geostructures. In this paper, the efficiency and predictability of different artificial intelligence (AI)-based models in generating improved 3D spatial distributions of the BL for an area in Stockholm, Sweden, were explored. Multilayer percepterons, generalized feed-forward neural network (GFFN), radial based function, and support vector regression (SVR) were developed and compared to ordinary kriging geostatistical technique. Analysis of the improvement in progress using confusion matrixes showed that the GFFN and SVR provided closer results to realities. The ranking of performance accuracy using different statistical errors and precision–recall curves also demonstrated the superiority and robustness of the GFFN and SVR compared to the other models. The results indicated that in the absence of measured data the AI models are flexible and efficient tools in creating more accurate spatial 3D models. Analyses of confidence intervals and prediction intervals confirmed that the developed AI models can overcome the associated uncertainties and provide appropriate prediction at any point in the subsurface of the study area.

Prediction of bed load sediments using different artificial neural network models

Modeling and prediction of bed loads is an important but difficult issue in river engineering. The introduced empirical equations due to restricted applicability even in similar conditions provide different accuracies with each other and measured data. In this paper, three different artificial neural networks (ANNs) including multilayer percepterons, radial based function (RBF), and generalized feed forward neural network using five dominant parameters of bed load transport formulas for the Main Fork Red River in Idaho-USA were developed. The optimum models were found through 102 data sets of flow discharge, flow velocity, water surface slopes, flow depth, and mean grain size. The deficiency of empirical equations for this river by conducted comparison between measured and predicted values was approved where the ANN models presented more consistence and closer estimation to observed data. The coefficient of determination between measured and predicted values for empirical equations varied from 0.10 to 0.21 against the 0.93 to 0.98 in ANN models. The accuracy performance of all models was evaluated and interpreted using different statistical error criteria, analytical graphs and confusion matrixes. Although the ANN models predicted compatible outputs but the RBF with 79% correct classification rate corresponding to 0.191 network error was outperform than others.

Viscosity Prediction of Lubricants by a General Feed-Forward Neural Network.

Modern industrial lubricants are often blended with an assortment of chemical additives to improve the performance of the base stock. Machine learning-based predictive models allow fast and veracious derivation of material properties and facilitate novel and innovative material designs. In this study, we outline the design and training process of a general feed-forward artificial neural network that accurately predicts the dynamic viscosity of oil-based lubricant formulations. The network hyperparameters are systematically optimized by Bayesian optimization, and strongly correlated/collinear features are trimmed from the model. By harnessing domain knowledge in the selection of features, the quantitative structure-property relationship model is built with a relatively simple feature set and is versatile in predicting the dynamic viscosity of lubricant oils with and without enhancement by viscosity modifiers (VMs). Moreover, partial dependency, local-interpretable model-agnostic explanations, and Shapley values consistently show that the eccentricity index, Crippen MR, and Petitjean number are important predictors of viscosity. All in all, the neural model is reasonably accurate in predicting the dynamic viscosity of lubricant solvents and VM-enhanced lubricants with an R2 of 0.980 and 0.963, respectively.

Forecast the Influenza Pandemic Using Machine Learning

Forecasting future outbreaks can help in minimizing their spread. Influenza is a disease primarily found in animals but transferred to humans through pigs. In 1918, influenza became a pandemic and spread rapidly all over the world becoming the cause behind killing one-third of the human population and killing one-fourth of the pig population. Afterwards, that influenza became a pandemic several times on a local and global levels. In 2009, influenza ‘A’ subtype H1N1 again took many human lives. The disease spread like in a pandemic quickly. This paper proposes a forecasting modeling system for the influenza pandemic using a feed-forward propagation neural network (MSDII-FFNN). This model helps us predict the outbreak, and determines which type of influenza becomes a pandemic, as well as which geographical area is infected. Data collection for the model is done by using IoT devices. This model is divided into 2 phases: The training phase and the validation phase, both being connected through the cloud. In the training phase, the model is trained using FFNN and is updated on the cloud. In the validation phase, whenever the input is submitted through the IoT devices, the system model is updated through the cloud and predicts the pandemic alert. In our dataset, the data is divided into an 85% training ratio and a 15% validation ratio. By applying the proposed model to our dataset, the predicted output precision is 90%.

Prediction of Uniaxial Compressive Strength of Different Quarried Rocks Using Metaheuristic Algorithm

The direct measurement of uniaxial compressive strength (UCS) as one of the main important rock engineering parameters is destructive, cumbersome, difficult and costly. Therefore, the prediction of this parameter using simpler, cheaper indirect methods is of interest. In this paper, the UCS was predicted using a developed hybrid intelligent model including generalized feedforward neural network (GFFN) incorporated with imperialist competitive algorithm (ICA). To find the optimum model, 197 sets including rock class, density, porosity, P-wave velocity, point load index and water absorption from almost all over quarries of Iran were compiled. The efficiency and performance of GFFN and hybrid ICA-GFFN models subjected to different error criteria and established confusion matrixes were compared to multilayer perceptron (MLP) and radial basis function (RBF) neural network models as well as conducted multivariate regression. The hybrid ICA-GFFN with 11.37%, 14.27% and 22.74% improvement in correct classification rate over than GFFN, RBF and MLP demonstrated superior predictability level. The results indicated that the developed ICA-GFFN model as a feasible and accurate enough tool can effectively be applied for UCS prediction purposes. Using the sensitivity analyses, the P-wave velocity and rock class were identified as the most and least influences factors on predicted UCS.

An artificial neural network based model to predict spatial soil type distribution using piezocone penetration test data (CPTu)

Soil types mapping and the spatial variation of soil classes are essential concerns in both geotechnical and geoenvironmental engineering. Because conventional soil mapping systems are time-consuming and costly, alternative quick and cheap but accurate methods need to be developed. In this paper, a new optimized multi-output generalized feed forward neural network (GFNN) structure using 58 piezocone penetration test points (CPTu) for producing a digital soil types map in the southwest of Sweden is developed. The introduced GFNN architecture is supported by a generalized shunting neuron (GSN) model computing unit to increase the capability of nonlinear boundaries of classified patterns. The comparison conducted between known soil type classification charts, CPTu interpreting procedures, and the outcomes of the GFNN model indicates acceptable accuracy in estimating complex soil types. The results show that the predictability of the GFNN system offers a valuable tool for the purpose of soil type pattern classifications and providing soil profiles.

Optimized developed artificial neural network-based models to predict the blast-induced ground vibration

Blasting has been widely used as an accepted mechanism in mining, construction, and rock engineering projects. However, inappropriate control of the blasting-induced ground vibration as an inevitable side effect can cause severe problem for the nearby areas. Therefore, developing the predictive models to estimate the blasting-induced ground vibrations can be considered as an attractive practical issue in engineering projects both in designing and operational stages. In the present paper, blasting-induced ground vibration at Masjed Soleyman earth dam in southwest of Iran in terms of peak particle velocity (PPV) using two different artificial neural network (ANN)-based models has been assessed and predicted. The multilayer perceptron (MLP) and generalized feed forward neural network (GFNN) were developed and optimized using monitored blast records. The total charge, charge per delay, and distance from blasting point were the input parameters. The quality and performance of introduced ANN topologies were compared to known conventional empirical predictors and then examined by different statistical indices and sensitivity analyses criteria. Although both GFNN and MLP indicated higher degree of safety and reliability in prediction of PPV, but the validation process using the unseen randomized data highlighted better performance and more accuracy in GFNN model with respect to MLP and common empirical predictors. Therefore, the GFNN with 3-4-3-1 structure and R2 = 0.954 between the measured and predicted PPV values was recognized as the optimized developed structure for the studied area.

Human Age and Gender Prediction Based on Neural Networks and Three Sigma Control Limits

ABSTRACTA person’s face provides a lot of information such as age, gender, and identity. Faces play an important role in the estimation/prediction of the age and gender of persons, just by looking at their face. Perceiving human faces and modeling the distinctive features of human faces that contribute most toward face recognition are some of the challenges faced by computer vision and psychophysics researchers. There are many methods have been proposed in the literature for the facial features for age and gender classification. In this research, an attempt is made to classify human age and gender using feed forward propagation neural networks in coarser level. Further final classification is done using 3-sigma control limits in finer level. Proposed approach efficiently classifies three age groups including children, middle-aged adults, and old-aged adults. Similarly two gender groups classified into male and female by the proposed method.The performance of the system is further improved by employing multiple hierarchical decision using three sigma control limits applied on the output of the neural network classifier. The mean and standard deviation has been considered on the output generated from the neural network classifier, and three sigma control limits has been applied to define the range of values for the specific category of age and gender. The efficiency of the system is demonstrated through the experimental results using benchmark database images.

Thermal Performance of Flat Tube Louver Fin Heat Exchanger using Artificial Neural Network

In this paper, an application of artificial neural networks (ANNs) is presented to predict the outlet temperatures on both fluid side of flat tube louver fin type heat exchanger used in flat tube louver fin heat exchanger. A validated numerical code is developed using MATLAB to generate huge data sets which gives outlet temperature of both the fluids. Three-layer feed-forward back propagation neural network is developed to model the thermal performance of heat exchanger. The ANN is thereafter trained using the numerical data to predict the outlet temperatures in heat exchanger. ANN configurations are set by selecting the number of hidden layers and nodes to achieve error convergence. Different network configurations are also examined for searching a better network for prediction. Keywords: Artificial Neural Networks, louver fin, radiator, numerical code, compact heat exchanger Cite this Article Manish K. Rathod. Thermal Performance of Flat Tube Louver Fin Heat Exchanger using Artificial Neural Network. Journal of Thermal Engineering and Applications . 2017; 4(3): 6–12p.

A Comparative Study of Various Hybrid Wavelet Feedforward Neural Network Models for Runoff Forecasting

Considering network topologies and structures of the artificial neural network (ANN) used in the field of hydrology, one can categorize them into two different generic types: feedforward and feedback (recurrent) networks. Different types of feedforward and recurrent ANNs are available, but multilayer perceptron type of feedforward ANN is most commonly used in hydrology for the development of wavelet coupled neural network (WNN) models. This study is conducted to compare performance of the various wavelet based feedforward artificial neural network (ANN) models. The feedforward ANN types used in the study include the multilayer perceptron neural network (MLPNN), generalized feedforward neural network (GFFNN), radial basis function neural network (RBFNN), modular neural network (MNN) and neuro-fuzzy neural network (NFNN) models. The rainfall-runoff data of four catchments located in different hydro-climatic regions of the world is used in the study. The discrete wavelet transformation (DWT) is used in the present study to decompose input rainfall data using db8 wavelet function. A total of 220 models are developed in this study to evaluate the performance of various feedforward neural network models. Performance of the developed WNN models is compared with their counterpart simple models developed without applying wavelet transformation (WT). The results of the study are further compared with - multiple linear regression (MLR) model which suggest that the WNN models outperformed their counterpart simple models. The hybrid wavelet models developed using MLPNN, the GFFNN and the MNN models performed best among the six selected data driven models explored in the study. Moreover, performance of the three best models is found to be similar and thus the hybrid wavelet GFFNN and the MNN models can be considered as an alternative to the most commonly used hybrid WNN models developed using MLPNN. The study further reveals that the wavelet coupled models outperformed their counterpart simple models only with the parsimonious input vector.

Error estimation in high-throughput density functional theory calculation for material property: elastic constants of cubic binary alloy case

Estimating Density Functional Theory (DFT) calculation error is an important while challenging task in computational material science. The calculation contains inherent errors due to improper input parameters and approximated exchange-correlation functional. In this paper, we present a data-driven approach of using machine learning techniques to estimate the error of DFT calculation. We prepare the data by high-throughput first principle DFT simulation and experimental data collection. The single-hidden layer back propagation feedforward neural network (SLBPFN) constructed based on the proposed cross validation algorithm, and support vector machine for regression (SVR) techniques are employed to build regression models to predict the DFT calculation error. As a demonstration, the developed regression models are used to predict errors in calculating elastic constants of cubic binary alloys. It has been demonstrated that the machine learning techniques can predict DFT calculation error of elastic constants with an acceptable accuracy. It also shows the BP neural network built by our proposed cross validation algorithm can provide a better prediction. Our study is a first-invasive work of using machine learning techniques to estimate the errors in calculating elastic constants of binary alloys.