Neural Network Programming Research Articles

An improved large-signal model by artificial neural network for the MOSFETs operating in the breakdown region

An improved large-signal breakdown model establishment applying an artificial neural network (ANN) approach is presented for metal-oxide-semiconductor field-effect transistors (MOSFETs) operating in the breakdown regime for the first time. A neural network program with the feed-forward back propagation algorithm and Levenberg Marquardt optimization is utilized to obtain the MOSFET large-signal model for breakdown operation. When compared with the conventional model without the breakdown effects considered, more accurate large-signal characteristics in the breakdown regime can be obtained by incorporating the avalanche network using the ANN approach. The breakdown network is demonstrated to be significant for large-signal characterization at high bias according to the analysis of minimum acceptable error. Besides, the divergence problem due to the neglected avalanche network can be avoided during ANN training in the avalanche regime. The accuracy of the presented model can keep about 2% error. The presented ANN approach can be applied to device large-signal modeling in the impact ionization regime.

Estimating scour below inverted siphon structures using stochastic and soft computing approaches

This paper uses nonlinear regression, Artificial Neural Network (ANN) and Genetic Programming (GP) approaches for predicting an important tangible issue i.e. scours dimensions downstream of inverted siphon structures. Dimensional analysis and nonlinear regression-based equations was proposed for estimation of maximum scour depth, location of the scour hole, location and height of the dune downstream of the structures. In addition, The GP-based formulation results are compared with experimental results and other accurate equations. The results analysis showed that the equations derived from Forward Stepwise nonlinear regression method have correlation coefficient of R2=0.962 , 0.971 and 0.991 respectively. This correlates the relative parameter of maximum scour depth (s/z) in comparison with the genetic programming (GP) model and artificial neural network (ANN) model. Furthermore, the slope of the fitted line extracted from computations and observations for dimensionless parameters generally presents a new achievement for sediment engineering and scientific community, indicating the superiority of artificial neural network (ANN) model

An analog neural network approach for the least absolute shrinkage and selection operator problem

This paper addresses the analog optimization for non-differential functions. The Lagrange programming neural network (LPNN) approach provides us a systematic way to build analog neural networks for handling constrained optimization problems. However, its drawback is that it cannot handle non-differentiable functions. In compressive sampling, one of the optimization problems is least absolute shrinkage and selection operator (LASSO), where the constraint is non-differentiable. This paper considers the hidden state concept from the local competition algorithm to formulate an analog model for the LASSO problem. Hence, the non-differentiable limitation of LPNN can be overcome. Under some conditions, at equilibrium, the network leads to the optimal solution of the LASSO. Also, we prove that these equilibrium points are stable. Simulation study illustrates that the proposed analog model and the traditional digital method have the similar mean squared performance.

Accuracy of rainfall estimates at high altitude in the Garhwal Himalaya (India): A comparison of secondary precipitation products and station rainfall measurements

Accurate estimation of the magnitude and spatio-temporal variability of rainfall in the Indian Himalaya is difficult because of the sparse and limited network of ground stations located within complex terrain, as well as the difficulty of maintaining the stations over time. Thus, secondary rainfall sources are important to hydrological and hazard studies, if they reproduce the dynamics of rainfall satisfactorily. In this work, we evaluate four secondary products in the Garhwal Himalaya in India, with a focus on their application within the Mandakini River Catchment, the site of a devastating flood and multiple large landslides in 2013. The analysis included two satellite products: from the Tropical Rainfall Measuring Mission (TRMM) and the Precipitation Estimation from Remotely Sensed Information using Artificial Neural Networks (PERSIANN) program, as well as two gridded products: the Asian Precipitation Highly Resolved Observational Data Integration Towards Evaluation of Water Resources (APHRODITE) product and the India Meteorological Department (IMD) product. In comparing the four products against data collected at four ground stations (Rudraprayag, Joshimath, Purola, and Mukhim) using a variety of statistical indices, we determined that the IMD and TRMM products were superior to the others. In particular, the IMD product ranked the best for most indices including probability of detection (POD), false alarm ratio (FAR), receiver operating curve (ROC), and root mean squared error (RMSE). The TRMM product performed satisfactorily in terms of bias and detecting daily maximum monsoon rainfall at three of the four stations. The APHRODITE product had POD, FAR and ROC values that were among the best at higher rainfall depths at the Mukhim station. The PERSIANN product generally did not perform well based on these indices, consistently underestimating station rainfall depths. Finally, the IMD product could document the daily rainfall distribution during the June 2013 flood in the Mandakini Catchment and adjoining places.

Comparing the Efficiency of Artificial Neural Network and Gene Expression Programming in Predicting Coronary Artery Disease

Background: Angiography, as the gold standard for the diagnosis of coronary artery disease, has made an attempt to predict coronary artery disease by comparing the efficiency of gene expression programming, as a new data mining technique, and artificial neural network, as a conventional technique. Besides, the study went further to present the results of feature selection based on stepwise backward elimination, classification and regression tree. Methods: The subjects were assessed for nine coronary artery disease risk factors to develop a prediction model for the disease. They included 13,288 patients who were chosen to undergo angiography for the diagnosis of coronary artery disease; from this sample, 4059 subjects were free from the disease while 9169 were suffering from it. Modeling was carried out based on gene expression programming and artificial neural network techniques. The Delong’s test was then used to choose the final model based on the area under the Receiver Operating Characteristic (ROC) curve. Results: The model, developed based on artificial neural network, had AUC of 0.719, accuracy of 73.39%, sensitivity of 93.44% and specificity of 28.34%. On the other hand, the model, formulated based on gene expression programming, had AUC of 0.720, accuracy of 73.94%, sensitivity of 93.29% and specificity of 31.43%. Delong’s test showed no significant difference between the two models (p value=0/789). Then, feature selection method was used to choose a model with four risk factors and an accuracy rate of 73.26%. Conclusion: Comparison of the results showed no significant difference between the two modeling techniques. The gene expression programming model was very easy to present and interpret; it could also be easily converted to other programming languages; so, with these features in mind, the researchers preferred to choose this technique.

進化的計算手法を用いた論理ゲートネットワークの生成に関する研究

In recent years, various techniques, such as Neural Network, Deep Learning and Genetic Programming(GP), to realize the adaptive behavior of the robot have been proposed. This paper aims to construct the control structure of the robot moving in the real world to realize the adaptive behavior through Evolutionary Hardware approach. The proposal method mentioned in this paper tries to build a gate network consisted of NAND gates by using Cellular Automaton and the adequate rule set of the Cellular Automaton is obtained by Genetic Algorithm(GA). And it is applied to the position control problem of the two-linked manipulator to confirm the effectiveness.

Predicting discharge coefficient of triangular labyrinth weir using extreme learning machine, artificial neural network and genetic programming

Weirs are a type of hydraulic structure used to direct and transfer water flows in the canals and overflows in the dams. The important index in computing flow discharge over the weir is discharge coefficient (Cd). The aim of this study is accurate determination of the Cd in triangular labyrinth side weirs by applying three intelligence models [i.e., artificial neural network (ANN), genetic programming (GP) and extreme learning machine (ELM)]. The calculated discharge coefficients were then compared with some experimental results. In order to examine the accuracy of Cd predictions by ANN, GP and ELM methods, five statistical indices including coefficient of determination (R2), root-mean-square error (RMSE), mean absolute percentage error (MAPE), SI and δ have been used. Results showed that R2 values in the ELM, ANN and GP methods were 0.993, 0.886 and 0.884, respectively, at training stage and 0.971, 0.965 and 0.963, respectively, at test stage. The ELM method, having MAPE, RMSE, SI and δ values of 0.81, 0.0059, 0.0082 and 0.81, respectively, at the training stage and 0.89, 0.0063, 0.0089 and 0.88, respectively, at the test stage, was superior to ANN and GP methods. The ANN model ranked next to the ELM model.

Earthquake magnitude prediction in Hindukush region using machine learning techniques

Earthquake magnitude prediction for Hindukush region has been carried out in this research using the temporal sequence of historic seismic activities in combination with the machine learning classifiers. Prediction has been made on the basis of mathematically calculated eight seismic indicators using the earthquake catalog of the region. These parameters are based on the well-known geophysical facts of Gutenberg–Richter’s inverse law, distribution of characteristic earthquake magnitudes and seismic quiescence. In this research, four machine learning techniques including pattern recognition neural network, recurrent neural network, random forest and linear programming boost ensemble classifier are separately applied to model relationships between calculated seismic parameters and future earthquake occurrences. The problem is formulated as a binary classification task and predictions are made for earthquakes of magnitude greater than or equal to 5.5 ( $$M \ge$$ 5.5), for the duration of 1 month. Furthermore, the analysis of earthquake prediction results is carried out for every machine learning classifier in terms of sensitivity, specificity, true and false predictive values. Accuracy is another performance measure considered for analyzing the results. Earthquake magnitude prediction for the Hindukush using these aforementioned techniques show significant and encouraging results, thus constituting a step forward toward the final robust prediction mechanism which is not available so far.

Small-signal modeling for p-n junctions in the breakdown region at different temperatures using artificial neural networks

In this paper, an artificial neural network (ANN) approach is applied to efficiently and accurately determine small-signal parameters of a breakdown model for silicon p-n junctions operating in the impact ionization region at different temperatures for the first time. A feed-forward back propagation neural network program with a Levenberg Marquardt optimization algorithm is employed to implement this ANN-based parameter determination approach. Measured S-parameters at different temperatures which are regarded as a training data-set are fitted by this program to determine small-signal parameters of the breakdown equivalent circuit model at different temperatures. Multiplication factors of p-n junctions at different temperatures are also extracted. This simple and accurate breakdown characterization method based on the ANNs can be applicable to automatic parameter determination for devices operating in the breakdown region.

Recurrent Cartesian Genetic Programming of Artificial Neural Networks

Cartesian Genetic Programming of Artificial Neural Networks is a NeuroEvolutionary method based on Cartesian Genetic Programming. Cartesian Genetic Programming has recently been extended to allow recurrent connections. This work investigates applying the same recurrent extension to Cartesian Genetic Programming of Artificial Neural Networks in order to allow the evolution of recurrent neural networks. The new Recurrent Cartesian Genetic Programming of Artificial Neural Networks method is applied to the domain of series forecasting where it is shown to significantly outperform all standard forecasting techniques used for comparison including autoregressive integrated moving average and multilayer perceptrons. An ablation study is also performed isolating which specific aspects of Recurrent Cartesian Genetic Programming of Artificial Neural Networks contribute to it’s effectiveness for series forecasting.

Lagrange Programming Neural Network for Nondifferentiable Optimization Problems in Sparse Approximation.

The major limitation of the Lagrange programming neural network (LPNN) approach is that the objective function and the constraints should be twice differentiable. Since sparse approximation involves nondifferentiable functions, the original LPNN approach is not suitable for recovering sparse signals. This paper proposes a new formulation of the LPNN approach based on the concept of the locally competitive algorithm (LCA). Unlike the classical LCA approach which is able to solve unconstrained optimization problems only, the proposed LPNN approach is able to solve the constrained optimization problems. Two problems in sparse approximation are considered. They are basis pursuit (BP) and constrained BP denoise (CBPDN). We propose two LPNN models, namely, BP-LPNN and CBPDN-LPNN, to solve these two problems. For these two models, we show that the equilibrium points of the models are the optimal solutions of the two problems, and that the optimal solutions of the two problems are the equilibrium points of the two models. Besides, the equilibrium points are stable. Simulations are carried out to verify the effectiveness of these two LPNN models.

Modeling scour depth downstream of grade-control structures using data driven and empirical approaches

Local scour occurs in the immediate vicinity of structures as a result of impinging on a bed with a high velocity flow. Prediction of scour depth has an important role in control structure management and water resource engineering issues, so a study of new heuristic expressions governing it is necessary. The present study aims to investigate different methods' capabilities to estimate scour depth downstream of grade-control structures using field measurements from the literature. Accordingly, data driven feed forward neural network and gene expression programming techniques were selected for the investigation. Additionally, the optimum data driven based scour depth models were compared with the corresponding physical–empirical based formulas. Three data categories corresponding to (a) scouring downstream of a ski-jump bucket, (b) a sharp-crested weir, and (c) an inclined slope controlled structure (as grade-control structures) were applied as reference patterns for developing and validating the applied models. A sensitivity analysis was also performed to identify the most influential parameters on scouring. The obtained results indicated that the applied methods have promising performance in estimating the scour depth downstream of spillways and control structures. Nevertheless, the applied data driven approaches show higher accuracy than the corresponding traditional formulas.

Application of Multivariate Adaptive Regression Spline-Assisted Objective Function on Optimization of Heat Transfer Rate Around a Cylinder

Abstract The present study aims to predict the heat transfer characteristics around a square cylinder with different corner radii using multivariate adaptive regression splines (MARS). Further, the MARS-generated objective function is optimized by particle swarm optimization. The data for the prediction are taken from the recently published article by the present authors [P. Dey, A. Sarkar, A.K. Das, Development of GEP and ANN model to predict the unsteady forced convection over a cylinder, Neural Comput. Appl. (2015) 1–13]. Further, the MARS model is compared with artificial neural network and gene expression programming. It has been found that the MARS model is very efficient in predicting the heat transfer characteristics. It has also been found that MARS is more efficient than artificial neural network and gene expression programming in predicting the forced convection data, and also particle swarm optimization can efficiently optimize the heat transfer rate.

Soil temperature modeling at different depths using neuro-fuzzy, neural network, and genetic programming techniques

The applicability of artificial neural networks (ANN), adaptive neuro-fuzzy inference system (ANFIS), and genetic programming (GP) techniques in estimating soil temperatures (ST) at different depths is investigated in this study. Weather data from two stations, Mersin and Adana, Turkey, were used as inputs to the applied models in order to model monthly STs. The first part of the study focused on comparison of ANN, ANFIS, and GP models in modeling ST of two stations at the depths of 10, 50, and 100 cm. GP was found to perform better than the ANN and ANFIS-SC in estimating monthly ST. The effect of periodicity (month of the year) on models’ accuracy was also investigated. Including periodicity component in models’ inputs considerably increased their accuracies. The root mean square error (RMSE) of ANN models was respectively decreased by 34 and 27 % for the depths of 10 and 100 cm adding the periodicity input. In the second part of the study, the accuracies of the ANN, ANFIS, and GP models were compared in estimating ST of Mersin Station using the climatic data of Adana Station. The ANN models generally performed better than the ANFIS-SC and GP in modeling ST of Mersin Station without local climatic inputs.

Application of a genetic algorithm in predicting the percentage of shear force carried by walls in smooth rectangular channels

Shear stress comprises basic information for predicting average depth velocity and discharge in channels. With knowledge of the percentage of shear force carried by walls (%SFw) it is possible to more accurately estimate shear stress values. The %SFw in smooth rectangular channels was predicted by extending two soft computing methods: Genetic Algorithm Artificial (GAA) neural network and Genetic Programming (GP). In order to investigate the percentage of shear force, 8 data series with a total of 69 different data were used. The outcomes of the GAA model (an equation) and the GP model (a program) were presented. In order to detect these models’ ability to predict %SFw, the obtained results were compared with several equations derived by other researchers. The GAA model with RMSE of 2.5454 and the GP model with RMSE of 3.0559 performed better than other equations with mean RMSE of about 9.630.

Lagrange Programming Neural Network Approach for Target Localization in Distributed MIMO Radar

In this paper, the problem of source localization in distributed multiple-input multiple-output (MIMO) radar using bistatic range measurements, which correspond to the sum of transmitter-to-target and target-to-receiver distances, is addressed. Our solution is based on the Lagrange programming neural network (LPNN), which is an analog neural computational technique for solving nonlinear constrained optimization problems according to the Lagrange multiplier theory. The local stability of the proposed positioning algorithm is also investigated. Furthermore, we have extended the LPNN based approach to more challenging scenarios, namely, when time synchronization among all antennas cannot be fulfilled, and there are position uncertainties in the MIMO radar transmit and receive elements. The optimality of the developed algorithms is demonstrated by comparing with the Cramer-Rao lower bound via computer simulations.

Improved artificial neural network RF noise model for MOSFETs operating in avalanche region

An artificial neural network (ANN)-based noise model of MOSFETs for breakdown operation is presented for the first time. This ANN architecture for avalanche noise model determination is developed using a feed-forward back propagation neural network program with optimisation algorithm of Levenberg Marquardt. Good agreement between measured and simulated noise parameters is achieved in the breakdown region. This simple and accurate breakdown characterisation method based on the ANN can be applicable to noise modelling for devices operating in the breakdown region.

Prediction of Compressive Strength of Concrete Using Artificial Neural Network and Genetic Programming

An effort has been made to develop concrete compressive strength prediction models with the help of two emerging data mining techniques, namely, Artificial Neural Networks (ANNs) and Genetic Programming (GP). The data for analysis and model development was collected at 28-, 56-, and 91-day curing periods through experiments conducted in the laboratory under standard controlled conditions. The developed models have also been tested on in situ concrete data taken from literature. A comparison of the prediction results obtained using both the models is presented and it can be inferred that the ANN model with the training function Levenberg-Marquardt (LM) for the prediction of concrete compressive strength is the best prediction tool.

Development and analysis of neural networks to predict the efficiency parameters of regenerator checker of glass furnace

The basic modern types of checkers and refractory materials for regenerative heat exchangers of glass furnaces were described. Operating features of using refractory materials of the checker depending on the purpose and the height of the regenerator design were given. To solve the inverse problem of predicting the regenerator parameters and classifying the refractory material of the checker depending on the coolant temperature at the checker flue outlet, the methods of neural network programming were applied, and a neural network based on multi-layer perceptron was created. The structure of this neural network was analyzed, the patterns of using different types of activation functions to solve various prediction problems were identified. The advantages of neural network models for the successful solution of prediction problems and classification of parameters of regenerative heat exchangers compared to existing finite-difference methods used for solving non-stationary problems of complex heat transfer in the checker were revealed.

Modeling deformation modulus of a stratified sedimentary rock mass using neural network, fuzzy inference and genetic programming

This paper investigates a series of experimental results and numerical simulations employed to estimate the deformation modulus of a stratified rock mass. The deformation modulus of rock mass has a significant importance for some applications in engineering geology and geotechnical projects including foundation, slope, and tunnel designs. Deformation modulus of a rock mass can be determined using large scale in-situ tests. This large scale sophisticated in-situ testing equipments are sometimes difficult to install, plus time consuming to be employed in the field. Therefore, this study aims to estimate indirectly the deformation modulus values via empirical methods such as the neural network, neuro fuzzy and genetic programming approaches. A series of analyses have been developed for correlating various relationships between the deformation modulus of rock mass, rock mass rating, rock quality designation, uniaxial compressive strength, and elasticity modulus of intact rock parameters. The performance capacities of proposed models are assessed and found as quite satisfactory. At the completion of a comparative study on the accuracy of models, in the results, it is seen that overall genetic programming models yielded more precise results than neural network and neuro fuzzy models.