Counterpropagation Neural Network Research Articles

Random forest and counter propagation neural network model used to study the consequence of nano silica on the mechanical performance of high strength concrete at elevated temperature

Random forest and counter propagation neural network model used to study the consequence of nano silica on the mechanical performance of high strength concrete at elevated temperature

Analysis of mechanical properties of fly ash and bauxite residue based geopolymer concrete using ANN, Random Forest and Counter propagation neural network

Analysis of mechanical properties of fly ash and bauxite residue based geopolymer concrete using ANN, Random Forest and Counter propagation neural network

Merging Counter-Propagation and Back-Propagation Algorithms: Overcoming the Limitations of Counter-Propagation Neural Network Models.

Artificial neural networks (ANNs) are nowadays applied as the most efficient methods in the majority of machine learning approaches, including data-driven modeling for assessment of the toxicity of chemicals. We developed a combined neural network methodology that can be used in the scope of new approach methodologies (NAMs) assessing chemical or drug toxicity. Here, we present QSAR models for predicting the physical and biochemical properties of molecules of three different datasets: aqueous solubility, acute fish toxicity toward fat head minnow, and bio-concentration factors. A novel neural network modeling method is developed by combining two neural network algorithms, namely, the counter-propagation modeling strategy (CP-ANN) with the back-propagation-of-errors algorithm (BPE-ANN). The advantage is a short training time, robustness, and good interpretability through the initial CP-ANN part, while the extension with BPE-ANN improves the precision of predictions in the range between minimal and maximal property values of the training data, regardless of the number of neurons in both neural networks, either CP-ANN or BPE-ANN.

Performance Evaluation of Selected Classification Algorithms for Iris Recognition System

Quite a lot of techniques proven to be resourceful have been espoused to develop iris recognition system. Nearly hybridized, supervised and unsupervised artificial neural network techniques have been used individually in iris recognition system and other pattern recognitions but have not been compared based on some performance metrics. Counter Propagation Neural Network (CPNN) is a hybridized technique, Self-Organizing Feature Map (SOFM) is an unsupervised learning technique and Back Propagation Neural Network (BPNN) is a supervised learning technique. This research conducted a performance comparison of CPNN, SOFM and BPNN techniques to recognize iris dataset and establish the more efficient among the three techniques. A database of Three hundred (300) iris images was acquired from LAUIRIS dataset from LAUTECH Biometric Research Group database. The original images of 640*360 dimensions were resized to 200*200 without any alteration in the image using 80% for training and 20% for testing. Hough transform was applied to segment locate the iris region of eye image. Daugman’s Rubber Sheet Model was used to create a dimensionally consistent representation. Principle Component Analysis was applied for feature extraction and dimensionally reduction. Finally, classification and matching were done by using CPNN, SOFM and BPNN techniques. This was implemented using MATLAB (Matrix Laboratory) R2016b. The performance metrics used for classification were False Positive Rate (FPR), Sensitivity, Specificity, Re, cognition Accuracy and Recognition Time at 0.70 threshold value.The Recognition Accuracy (RA), Recognition Time (RT), False Acceptance Rate (FAR), Sensitivity and Specificity of the three selected techniques (CPNN, SOFM and BPNN) resulted in values of 95.17%, 177.48s, 6.33% and 93.67% for CPNN 92.50%, 179.69s, 9.00%, 94.00% and 90.99% for SOFM while BPNN had 91.17%, 187.88s, 10.33%, 92.67% and 89.67% respectively.This paper showed that CPNN classification technique performed best for iris recognition system in terms of RA and recognition time. This research output will serve as a basis to pre-inform and guide researchers in choosing an efficient kernel based feature extraction technique.

Functional stability of technological processes based on nonlinear dynamics with the application of neural networks

The processes of transformation of global information infrastructure and large-scale automation of production lead to the actual merger of automated production, data exchange and production technologies into a single self-regulatory system with minimal or no human intervention in the production process. Currently, there is a mass introduction of cyberphysical systems into production with the simultaneous application of the results obtained in the fields of artificial intelligence, robotics, the Internet of Things and so on. Implementing the goal of developing methods for organizing production processes of metal processing at machine-building enterprises using neural networks, the processes of global transformation of IT infrastructure were studied against the background of mass introduction of cyberphysical systems and breakthroughs in artificial intelligence and technological processes. The characteristics of the behavior of complex technical systems that implement the property of functional stability of such systems are studied. The processes of metal processing by cutting are characterized taking into account the peculiarities of the influence of deformation hardening, plastic deformations, self-oscillations and chaotic dynamics that occur in machining centers. Methods of application of neural networks in modeling of processes of mechanical processing of metals by cutting are described. A universal technique for constructing neural network models of the machining process on the basis of an artificial counter-propagation neural network is given. Based on the analysis, an intelligent system of analysis and forecasting of the dynamic stability of the technological process of cutting using parallel calculations, which guarantees the fulfillment of the necessary conditions to ensure the functional stability of the production process.

Application of Supervised SOM Algorithms in Predicting the Hepatotoxic Potential of Drugs.

The hepatotoxic potential of drugs is one of the main reasons why a number of drugs never reach the market or have to be withdrawn from the market. Therefore, the evaluation of the hepatotoxic potential of drugs is an important part of the drug development process. The aim of this work was to evaluate the relative abilities of different supervised self-organizing algorithms in classifying the hepatotoxic potential of drugs. Two modifications of standard counter-propagation training algorithms were proposed to achieve good separation of clusters on the self-organizing map. A series of optimizations were performed using genetic algorithm to select models developed with counter-propagation neural networks, X-Y fused networks, and the two newly proposed algorithms. The cluster separations achieved by the different algorithms were evaluated using a simple measure presented in this paper. Both proposed algorithms showed a better formation of clusters compared to the standard counter-propagation algorithm. The X-Y fused neural network confirmed its high ability to form well-separated clusters. Nevertheless, one of the proposed algorithms came close to its clustering results, which also resulted in a similar number of selected models.

Parametric design of turbocharger rotor system under exhaust emission loads via surrogate model

The common failure modes of a turbocharger rotor are bearing wear, impeller-casing rub events and shaft cracks due to high thermal currents. A reliable design of such rotors is therefore essential in order to avoid the vibration-based failure incidents during operation. The dynamic response gives vital information about the rotor condition as well as it helps in the inverse modelling to predict the optimum parameters. Present work deals with dynamic analysis and optimum design of a turbocharger rotor system subjected to ideal exhaust gas emission loads. The exhaust emission loads at the turbine in radial direction are modelled as Muszynska’s nonlinear seal forces, while the axial gas emission loads are treated as periodic blade passing excitations. The critical vibration characteristics of the rotor system running at uniform speed are analysed using the finite element model by considering nonlinear floating ring bearing forces and are initially validated with an experimental work performed on a turbocharger test-rig. Furthermore, the effects of gas load as well as floating ring bearing parameters on the critical frequencies and amplitudes are studied in detail. The effective input–output relationships are established using counter-propagation neural network (CPNN) learning scheme. Surrogate model using an improved Cuckoo search optimization in-conjunction with trained CPNN model in function evaluations is employed to predict the optimum values of the bearing parameters and relative emission load data. The relative output frequency response with optimized parameters is presented and effectiveness of the approach over the finite element based function evaluation scheme is illustrated.

A hybrid classical techniques and optimal decision model for iris recognition under variable image quality conditions

One of the best biometrics used for human verification and identification is iris recognition. The contrast of its unique characteristics differs from one candidate to another in which the iris pattern has numerous well-known features like uniqueness texture, stability and compactness representation for human identification. Among these facts, several approaches in these areas are localized, but there is still an abundant problems such as the low match rate of the score level and low accuracy. Therefore, a decision model should be essential for iris recognition frameworks. This paper proposes a iris recognition based on a decision model using a unified framework based on the integration of three detection schemes due to variation occurred in shading and position change using Smallest Univalue Segment Assimilating Nucleus (SUSAN), Generalized Hough Transform (GHT) and Viola–Jones (SUSANGHT-VJ) for eye detection, enhancing the dimmer and darker areas using fuzzy retinex method and normalizing the iris boundary using daugman’s Rubber Sheet Model for segmentation. Also, the iris corner points is extracted using Gabor Wavelet Transform (GWT), the vector properties of the blurred texture features are quantized using Local Phase Quantization (LPQ) and the optimal decision model based on Atom Search Optimization (ASO) and Feed Forward Counter propagation Neural Network (FFCNN) for matching score level and classification task. Furthermore, the current framework prevents false matches and inappropriate iris input, thus making the iris match score framework more reliable. The evaluation of the proposed approach is trained and tested with the employed eye template of iris and face datasets. Therefore, the results depict that the proposed technique gives a high recognition rate of 99.9% on different datasets compared to existing methods.

An optimal parameter identification approach in foil bearing supported high-speed turbocharger rotor system

High-speed turbocharger rotors are often supported on double film floating ring bearings. In recent times, foil bearings are increasingly used as an alternative to floating ring bearings for supporting such lightweight rotors. Present work aims at the dynamic modeling and stability studies of the automotive turbocharger rotor system supported on the airfoil bearings under varying operating conditions. The effects of different parameters including rotor spin speed, bearing clearance, bump foil thickness, and foil pitch on the dynamic response of the rotor system are illustrated. The nonlinear relationship between the bearing parameters and vibration response data is modeled using supervised neural network scheme. With the use of counter propagation neural network model, inverse identification methodology of the bearing parameters is performed. Further, an optimal design of foil bearing parameters using a surrogate model is proposed with a modified particle swarm optimization scheme. The resultant design parameters are employed, and the percentage reductions in vibration amplitudes of dynamic response are reported.

A Routing Protocol Based on CP Neural Network for Vehicular Ad Hoc Networks

It is important to design efficient routing protocols based on link connectivity and the selection of optimal link forwarding data in vehicular ad hoc networks (VANET). We propose A routing protocol based on counter propagation (CP) neural network for Vehicular Ad Hoc Networks (CP-GPSR), which is to take into account the connectivity probability of the link, adopts a routing method based on data packets forwarded at intersections. The priority of the next segment is classified by the CP neural network, vehicle select the next segment of highest priority for data forwarding at the intersection. The simulation results show that CP-GPSR has a high packet delivery rate, the control overhead is lower.

Development of a Plant Disease Classification System using an Improved Counter Propagation Neural Network

Development of a Plant Disease Classification System using an Improved Counter Propagation Neural Network

Counter-propagation Neural Network for Brain Tumor Classification

Brain tumor is a condition in which abnormal cells grow unnaturally in the brain. Depending on the size and type, the abnormal cells called tumors can be life-threatening if the patient does not take immediate treatment. The cause of tumor growth in the brain is the presence of risk factors such as family history and ionization radiation. Patients with brain tumors will experience several symptoms of a headache, nausea, memory loss, and changes in vision, speech, and hearing. Detection of brain tumors can be performed with the help of the medical device of Magnetic Resonance Imaging (MRI) Scan. Through the image of MRI Scan results, radiology specialists will interpret and analyze the brain condition. However, analysis and conclusions for this matter take a long period of time. Therefore, a method is required to classify the brain tumors through MRI images automatically. The method used in this research is Counter-propagation Neural Network. Prior to classification, the brain’s MRI image will be used as the input for the image pre-processing stage then go through the segmentation and feature extraction processes. Based on the test, it can be concluded that the proposed method can identify brain tumors with an accuracy of 92.5%.

Development of classification model based on neural networks for the process of iron ore beneficiation

The object of research is the processes of beneficiation of iron ore in the conditions of a mining and processing plant. Iron ore beneficiation factory near parallel to existing production lines or concentration sections. One of the key characteristics that determine the operating mode of the grinding apparatus is the crushing of ore, directly related to its strength. But unlike other parameters, the problem is with constant monitoring of the strength value. The determination of this parameter requires a laboratory study of the technological ore sample from the conveyor of the beneficiation section. The specifics of the working conditions of the beneficiation section complicate the monitoring of the strength parameter by installing a hardware sensor directly on the conveyor. Therefore, it is proposed to determine it by forecasting. Based on Big Data information technologies, using the accumulated statistical data, it is possible to forecast data between the technological samples.The technological process of ore beneficiation in the conditions of a mining and processing plant is systematically analyzed. The generalized structure of the classification model is presented, which, based on the accumulated statistical data of the beneficiation section based on the current parameters of the section, is able to determine the parameters of incoming raw materials. The unknown parameter is determined using the counterpropagation neural network, which combines the following algorithms: a self-organizing Kohonen map and a Grossberg star. Their combination leads to an increase in the generalizing properties of the network. The training sample is formed as a result of clustering the statistical data of the beneficiation section and selecting the cluster to which the current status of the section works.The presented forecasting algorithm, based on a combination of clustering methods and the use of a predictive neural network, allows the specialist to more quickly receive recommendations for making decisions regarding the behavior of the object compared to obtaining laboratory test data.

Machine learning to predict the specific optical rotations of chiral fluorinated molecules.

A chemoinformatics method was applied to the assignment of absolute configurations and to the quantitative prediction of specific optical rotations using a data set of 88 chiral fluorinated molecules (44 pairs of enantiomers). Counterpropagation neural networks were explored for the classification of enantiomers as dextrorotatory or levorotatory. Regression models were trained using multilayer perceptrons (MLP), random forests (RF) or multilinear regressions (MLR), on the basis of physicochemical atomic stereo (PAS) descriptors. New descriptors were also derived considering the common structural features of the data set (cPAS descriptors), which enabled RF models to predict the whole data set with R = 0.964, mean absolute error (MAE) of 9.8° and root mean square error (RMSE) of 12.5° in leave-one-pair-out cross-validation experiments. The predictions for the 30 compounds measured in chloroform were obtained with R = 0.971, MAE = 9.1° and RMSE = 12.5°, which compares favorably with quantum chemistry calculations reported in the literature.

Research on Intrusion Detection Model Based on improved CPN algorithm

In this paper, the clustering analysis method of intrusion detection technology is studied and analysed combined with the algorithm of Counter Propagation Networks. Based on the traditional Counter Propagation Neural Networks (CPN), this paper takes advantage of Gene Express Programming (GEP), optimize the input vector and α of the CPN network model. It overcomes the shortcoming of the network jitter, that because of the influence of the input vector on the connection weight of the Kohonen layer in the traditional CPN network algorithm. improves the convergence time of the network, and obtains a better network clustering effect, and verifies the effectiveness and superiority of the algorithm in the intrusion data analysis from the experiment.

Hybrid Design using Counter Propagation Neural Network-Genetic Algorithm Model for the Anomaly Detection in Online Transaction

In e-commerce, credit card fraud is an evolving challenge. The increase in the number of credit card transactions provides more opportunity for fraudsters to steal credit card numbers and execute fraud. Fraud detection is a continuously evolving discipline to tackle ever changing tactics to commit fraud. Existing fraud detection systems have not been so much efficient to reduce fraud transaction rate. Improvement in fraud detection practices has become essential to maintain existence of payment system. This research designed hybrid of Counter Propagation Neural Network and genetic algorithm (CPNN-GA) for the detection of anomaly in any online transactions.

Classification of Biomedical Images using Counter Propagation Neural Network

Classification of Biomedical Images using Counter Propagation Neural Network

NEURAL NETWORKS MODULE

It is considered a basic approach for hybrid neuron network creation. As an example, the counter propagation neural network is analyzed. It is effectively used for image processing. Two modes of this neuron network functioning are considered. They are: accreditation and interpolation. Interpolation mode permits to reveal more complex features and can supply more precise results. Based on this analysis it is developed a new hybrid structure that includes Kohonnen neural network and perceptron. It is proposed a learning algorithm of this hybrid neuron network.

Exploration of automatic learning to establish relationships between the molecular structure of chiral ionic liquids and the specific optical rotation

The experimental assignment of absolute configurations of chiral compounds is generally expensive and time consuming. Theoretical prediction of optical rotations can greatly assist in this endeavor. Herein a chemoinformatics exploration is reported to automatically distribute ionic liquids (IL) in a counterpropagation neural network incorporating a Kohonen layer for processing descriptors of cations, and five output layers to store optical rotations of ILs with five different anions. Cations were represented by chiral descriptors based on parities of the chiral centers assigned according to several steric and physicochemical atomic properties of the ligands. A data set with 106 chiral ionic liquids and their enantiomers was used covering a range of specific optical rotation between −147.3° and +147.3°. The obtained maps reveal relationships between the molecular structure of cations, the anions, and the optical rotation. The correct assignment of the sign of the specific optical rotations was achieved for 20 of the 22 pairs of enantiomers in an independent test set. CPG NNs estimated the value of the specific rotation with a RMS error of 22°, but more accurate predictions were obtained with a random forest that yielded a RMS error of 11°.

Building a smart lecture-recording system using MK-CPN network for heterogeneous data sources

Nowadays, lecture-recording systems play a vital role in collecting spoken discourse for e-learning. However, in view of the growing development of e-learning, the lack of content is becoming a problem. This research presents a smart lecture-recording (SLR) system that can record orations at the same level of quality as a human team, but with a reduced degree of human involvement. The proposed SLR system is composed of two subsystems, referred to as virtual cameraman (VC), and virtual director (VD), respectively. All camera man components of VC subsystem are automatic and can take actions that include target and event detection, tracking, and view searching. The videos taken by these three components are forwarded to the VD subsystem, in which the representative shot is chosen for recording or direct broadcasting. We refer to this function of the VD subsystem as shot selection that is based on the content analysis. The capability of shot selection is pre-trained through a machine-learning process characterized by the counter-propagation neural (CPN) network. However, the CPN network yielded poor results when the input data were heterogeneous data. To increases the accuracy of shot selection, we applied multiple kernel learning (MKL) techniques into CPN network, called MK-CPN, to transform all the heterogeneous data from different content analysis methods into unified space. A series of experiments for real lecture has been conducted. The results showed that the proposed SLR system can provide oration records close to some extend to those taken by real human teams. We believe that the proposed system may not be limited to live speeches, if it can be configured with appropriate training materials.