Search

Designing intelligent and robust autonomous navigation systems remains a great challenge in mobile robotics. Inverse reinforcement learning (IRL) offers an efficient learning technique from expert demonstrations to teach robots how to perform specific tasks without manually specifying the reward function. Most of existing IRL algorithms assume the expert policy to be optimal and deterministic, and are applied to experiments with relatively small-size state spaces. However, in autonomous navigation tasks, the state spaces are frequently large and demonstrations can hardly visit all the states. Meanwhile the expert policy may be non-optimal and stochastic. In this paper, we focus on IRL with large-scale and high-dimensional state spaces by introducing the neural network to generalize the expert’s behaviors to unvisited regions of the state space and an explicit policy representation is easily expressed by neural network, even for the stochastic expert policy. An efficient and convenient algorithm, Neural Inverse Reinforcement Learning (NIRL), is proposed. Experimental results on simulated autonomous navigation tasks show that a mobile robot using our approach can successfully navigate to the target position without colliding with unpredicted obstacles, largely reduce the learning time, and has a good generalization performance on undemonstrated states. Hence prove the robot intelligence of autonomous navigation transplanted from limited demonstrations to completely unknown tasks. [ABSTRACT FROM AUTHOR]

The information storage algorithm of a Hopfield-type network is a kind of self-correlation learning. This algorithm needs the condition in which vectors of learning patterns cross at right angles to each other. This paper proposes n method to make this condition easier. To this purpose, threshold is used. An investigation of how to decide quantity is included. This paper discusses a network with units whose state is 1 or 0. Fixing method of threshold has not been proposed before to this type of network. It is, therefore, proposed in this research; the goal is to facilitate learning patterns. [ABSTRACT FROM AUTHOR]

This paper proposes a fast learning algorithm of neural networks and evaluates the performances of adaptive equalizers using neural networks trained by the proposed algorithm in a frequency-selective fading channel. The backpropagation (BP) algorithm which is used widely to train neural networks has a slow convergence rate because it is based on the gradient descent method. This paper presents a fast learning algorithm using the recursive least squares (RLS) algorithm which has a fast convergence rate as an adaptive algorithm for adaptive linear filters. In the proposed algorithm, the sum of the squared error between the actual total input and the desired total input is used as the cost function to apply the RLS algorithm. A simulation result on the exclusive-OR problem indicates that the proposed algorithm is about 8.8 times faster than the BP for the number of iterations required to converge. Recently, there has been interest in adaptive equalizers as an application field of neural networks. However, the performance of an adaptive equalizer using a neural network in a frequency-selective fading channel which is observed in land mobile communications has never been evaluated. Therefore, in this paper, the performances of adaptive equalizers using neural networks trained by the proposed algorithm in a frequency-selective fading channel are evaluated. Especially, an adaptive equalizer using the selectively unsupervised learning neural network proposed by the authors is considered. The adaptive equalizer can reject the false learning by carrying out learning selectively. It is shown that the adaptive equalizer is superior to the conventional one and the one using the conventional neural network. [ABSTRACT FROM AUTHOR]

The traditional neural system which conducts the paired association has the real coupling coefficients and is not suited to the implementation on digital processing LSI because of its analog property. From such a viewpoint, this paper applies the traditional idea of calculating the coupling coefficients on the real field to the binary field and proposes a system which conducts the paired association using only the logic operation. More precisely, it is noted that the associative law applies to the matrix operation on the real field as well as the matrix operation for the binary code. The coupling coefficients for the binary code for the paired association arc derived by applying the sweeping-out method to the matrix of binary numbers. By this approach, a paired-associate system is obtained by a simpler training algorithm with easy LSI implementation. This paper describes first the traditional solution of the paired-associate problem on the real field and then proposes a method of solution based on that approach for the paired-associate problem on the binary field. The existence condition for the solution of the paired- associate problem is discussed for the proposed method. Finally, the noise-reduction power of the proposed method is estimated. [ABSTRACT FROM AUTHOR]

Copy number variation (CNV) is a non-negligible structural variation on the genome. And next-generation sequencing (NGS) technology is widely used to detect CNVs due to the feature of high throughput and low cost on the whole genome. Based on the original MFCNV method, this paper proposes an improved CNV detection method, which is called CNVABNN. In comparison to the MFCNV method, CNVABNN has three advantages: (1) It adds detectable categories, and refines the categories of loss into hemi_loss and homo_loss. (2) It utilizes the idea of integrated learning. The AdaBoost algorithm is used as the core framework and neural networks are used as weak classifiers, then CNVABNN combines all of the weak classifiers into a strong classifier. The overall performance of CNV detection is improved by using the strong classifier. (3) The detection is optimized by predicting CNVs twice through neural networks and voting mechanisms. To evaluate the performance of CNVABNN, six existing detection methods are used for comparison. The experimental results show that CNVABNN achieves better results in terms of precision, sensitivity, and F1-score for both simulated and real samples. [Display omitted] • A neural network computational model based on the AdaBoost framework for CNV detection is proposed. • Hemi_loss and homo_loss are introduced to make the classification of copy numbers realistic. • A two-stage prediction process improves the precision of detecting CNV. [ABSTRACT FROM AUTHOR]

This paper proposes a method for learning the process of curve fitting through a general class of totally positive rational bases. The approximation is achieved by finding suitable weights and control points to fit the given set of data points using a neural network and a training algorithm, called AdaMax algorithm, which is a first-order gradient-based stochastic optimization. The neural network presented in this paper is novel and based on a recent generalization of rational curves which inherit geometric properties and algorithms of the traditional rational Bézier curves. The neural network has been applied to different kinds of datasets and it has been compared with the traditional least-squares method to test its performance. The obtained results show that our method can generate a satisfactory approximation. [ABSTRACT FROM AUTHOR]

Radial Basis Function Neural Networks (RBFNs) are nowadays quite popular due to their ability to discover and approximate complex nonlinear dependencies within the data under analysis. Performance of the RBF network depends on numerous factors related to its initialization and training. The paper proposes an approach to the radial basis function networks design, where initial parameters of the network, output weights and parameters of the transfer function are set using the proposed agent-based population learning algorithm (PLA). The algorithm is validated experimentally. Advantages and main features of the PLA-based RBF designs are discussed basing on results of the computational experiment. [ABSTRACT FROM AUTHOR]

In this paper, we aim to solve the finite-horizon optimal control problem for a class of non-linear discrete-time switched systems using adaptive dynamic programming(ADP) algorithm. A new ε-optimal control scheme based on the iterative ADP algorithm is presented which makes the value function converge iteratively to the greatest lower bound of all value function indices within an error according to ε within finite time. Two neural networks are used as parametric structures to implement the iterative ADP algorithm with ε-error bound, which aim at approximating the value function and the control policy, respectively. And then, the optimal control policy is obtained. Finally, a simulation example is included to illustrate the applicability of the proposed method. [ABSTRACT FROM AUTHOR]

The utilization of back-propagation neural network in identification of internal dynamics of chaotic motion is found appropriate. However, during its training through Rumelhart algorithm, it is found that, a high learning rate (α) leads to rapid learning but the weights may oscillate, while a lower value of 'α' leads to slower learning process in weight updating formula Δvjk = αδjxi Momentum factor (μ) is to accelerate the convergence of error during the training in the equation wjk(t+1) = wjk(t) + αδkzj + μ{wjk(t) - wjk(t - 1)} vjk(t+1) = vjk(t) + αδkz + μ{vjk(t) - vjk(t-1) while transfer function sigmoid f(x) = 1/1+e-δx+n. It is the most complicated and experimental task to identify optimum value of 'α' and 'μ' during the training. To identiαy optimum value of 'α' and 'μ', firstly the network is trained with 10³ epochs under different values of 'α' in the close interval 0 < α <1 and μ = 1. At α = 0.3the convergence of initial weights and minimization of error (i.e., mean square error) process is found appropriate. Afterwards to find optimum value of μ, the network was trained again with α = 0.3 (fixed) and with different values of μ in the close interval 0 <μ < 1 for 10³ epochs. It was observed that the convergence of initial weights and minimization of error was appropriate with α = 0.3 and μ = 0.9. On this optimum value of α and μ the network was trained successfully from local minima of error = 1.67029292416874E-03 at 10³ epochs to global minima of error = 4.99180426869658E-04 at 15 105 epochs. At the global minima, the network has exhibited excellent performance in identification of internal dynamics of chaotic motion and in prediction of future values by past recorded data series. These essentials are presented through this research paper. [ABSTRACT FROM AUTHOR]

In this paper, an infinite-horizon optimal tracking control scheme is proposed for a class of nonlinear discrete-time switched systems. First, via system transformation, the optimal tracking problem is converted into designing an optimal regulator for the tracking error dynamics. And then, with convergence analysis in terms of value function and control policy, the iterative adaptive dynamic programming (ADP) algorithm is introduced to obtain the infinite-horizon optimal tracking controller which makes the value function close to its optimal value function. Next, two neural networks are used as parametric structures to implement the ADP algorithm, which aim at approximating the value function and the control policy, respectively. Finally, a simulation example is included to complement the theoretical discussions. [ABSTRACT FROM AUTHOR]

Prostate cancer is one of the leading causes of death by cancer among men in the world. Ultrasonography is said to be the safest technique in medical imaging so it is used extensively in prostate cancer detection. On the other hand, determining of prostate's boundary in TRUS (Transrectal Ultrasound) images is very necessary in lots of treatment methods prostate cancer. So the first and essential step for computer aided diagnosis (CAD) is the automatic prostate segmentation that is an open problem yet. But the low SNR, presence of strong speckle noise, Weak edges and shadow artifacts in these kinds of images limit the effectiveness of classical segmentation schemes. The classical segmentation methods fail completely or require post processing step to remove invalid object boundaries in the segmentation results. This paper has proposed a fully automatic algorithm for prostate segmentation in TRUS images that overcomes the explained problems completely. The presented algorithm contains three main stages. First, morphological smoothing and stick's filter are used for noise removing. A neural network is employed in the second step to find a point in prostate region. Finally in the last step, the prostate boundaries are extracted by GVF active contour. Some experiments for the performance validity of the presented method, compared with the extracted prostate by the proposed algorithm with manually-delineated boundaries by radiologist. The results show that our method extracts prostate boundaries with mean square area error lower than 4.4%. [ABSTRACT FROM AUTHOR]

In this paper, Levenberg-Marquardt feedforward MLP neural network (LMFFNN) was proposed to classify cervical cell images obtained from 100 patients including healthy, low-grade intraepithelial squamous lesion and high-grade intraepithelial squamous lesion cases. This neural network along with extracted cell image features is a new model for cervical cell image classification. The semiautomated cervical cancer diagnosis system is composed of two phases: image preprocessing/processing and feedforward MLP neural network. In the first stage, image preprocessing is done to reduce the existing noises without lowering the resolution. After that, image processing algorithms were applied to manually cropped cell images to achieve a linear plot which includes real components, were used as LMFFNN inputs for classification of cervical cell images. Based on the results, cervical cell images were classified successfully with 100 % correct classification rate using the proposed method. Moreover, the rates of sensitivity and specificity were calculated as 100 % using LMFFNN method. It was shown there was a good agreement between the expert decision and values gained from the ANN model. [ABSTRACT FROM AUTHOR]

To improve the accuracy of learning result, in practice multiple parties may collaborate through conducting joint Back-Propagation neural network learning on the union of their respective data sets. During this process no party wants to disclose her/his private data to others. Existing schemes supporting this kind of collaborative learning are either limited in the way of data partition or just consider two parties. There lacks a solution that allows two or more parties, each with an arbitrarily partitioned data set, to collaboratively conduct the learning. This paper solves this open problem by utilizing the power of cloud computing. In our proposed scheme, each party encrypts his/her private data locally and uploads the ciphertexts into the cloud. The cloud then executes most of the operations pertaining to the learning algorithms over ciphertexts without knowing the original private data. By securely offloading the expensive operations to the cloud, we keep the computation and communication costs on each party minimal and independent to the number of participants. To support flexible operations over ciphertexts, we adopt and tailor the BGN "doubly homomorphic" encryption algorithm for the multiparty setting. Numerical analysis and experiments on commodity cloud show that our scheme is secure, efficient, and accurate. [ABSTRACT FROM AUTHOR]

Abstract: This paper presents a novel approach of permeability prediction by combining cuckoo, particle swarm and imperialist competitive algorithms with Levenberg–Marquardt (LM) neural network algorithm in one of heterogeneous oil reservoirs in Iran. First, topology and parameters of the Artificial Neural Network (ANN) as decision variables were designed without the optimization method. Then, in order to improve the effectiveness of forecasting when ANN was applied to a permeability predicting problem, the design was performed using Cuckoo Optimization Algorithm (COA) algorithm. The validation test result from a new well data demonstrated that the trained COA–LM neural model can efficiently accomplish permeability prediction. Also, the comparison of COA with particle swarm optimization and imperialist competitive algorithms showed the superiority of COA on fast convergence and best optimum solution achievement. [Copyright &y& Elsevier]

This paper studies the monotonic type-2 fuzzy neural network (T2FNN), which can be adopted in many identification and prediction problems where the monotonicity property between the inputs and outputs is required. Sufficient conditions on the parameters of the T2FNN are first presented to ensure the monotonicity between the inputs and outputs. Then, data-driven design model for the monotonic T2FNN is built. Also, under the monotonicity constraints, a hybrid algorithm is provided to optimize the parameters of the monotonic T2FNN. This hybrid algorithm utilizes the constrained least squares method and the penalty function-based gradient descent algorithm to realize reasonable parameter initialization and optimization. At last, an application to the thermal comfort index prediction is given to verify the effectiveness of the monotonic T2FNN. Comparisons with other methods are also made. [ABSTRACT FROM AUTHOR]

Abstract: In solar PV (photovoltaic) system, tracking the module's MPP (maximum power point) is challenging due to varying climatic conditions. Moreover, the tracking algorithm becomes more complicated under the condition of partial shading due to the presence of multiple peaks in the power voltage characteristics. This paper presents a NN (neural network) based modified IC (incremental conductance) algorithm for MPPT (maximum power point tracking) in PV system. The PV system along with the proposed MPPT algorithm was simulated using Matlab/Simulink simscape tool box. The simulated system was evaluated under uniform and non-uniform irradiation conditions and the results are presented. For comparison, P&O (perturb and observe) and Fuzzy based Modified Hill Climbing algorithms were used for MPP tracking, and the results show that the proposed approach is effective in tracking the MPP under partial shading conditions. To validate the simulated system hardware implementation of the proposed algorithm was carried out using FPGA (Field Programmable Gate Array). [Copyright &y& Elsevier]

A novel vertical handover algorithm based on multi-attribute and neural network for heterogeneous integrated network is proposed in this paper. The whole frame of the algorithm is constructed by setting the network environment in which we use the network resources by switching between UMTS, GPRS, WLAN, 4G, and 5G. Each network build their own three-layer BP (Back Propagation, BP) neural network model and then the maximum transmission rate, minimum delay, SINR (signal to interference and noise ratio, SINR), bit error rate, user moving speed, and packet loss rate which can affect the overall performance of the wireless network are employed as reference objects to participate in the setting of BP neural network input layer neurons and the training and learning process of subsequent neural network data. Finally, the network download rate is adopted as prediction target to evaluate performance on the five wireless networks and then the vertical handover algorithm will select the right wireless network to perform vertical handover decision. The simulation results on MATLAB platform show that the vertical handover algorithm designed in this paper has a handover success rate up to 90% and realizes efficient handover and seamless connectivity between multi-heterogeneous networks. [ABSTRACT FROM AUTHOR]

This paper describes a very detailed and comprehensive description of a model predictive controller (MPC) applied to the buck converter working in Continuous Conduction Mode (CCM) to optimize the performance of the converter. The converter is designed for a fixed switching frequency of 1 MHz. The proposed model predictive control technique achieves improved set point tracking with minimal overshoot, offers consistent closed loop performance and has short settling time. Although MPC rejects unmeasured disturbances effectively, but it shows poor load and line regulation. Solution has been suggested in order to robustify the load and line regulation. Further neural network based MPC has also been designed. The impact of the MPC parameters such as prediction and control horizons, performance weights, and constraints on the performance is minutely observed. In order to investigate the effectiveness of the proposed algorithms, MPC is compared with the classical well-known PWM-PID controller. MATLAB/Simulink simulation results validate the control design procedure. [ABSTRACT FROM AUTHOR]

This paper proposes a wavelet-based fuzzy neuro control algorithm for the hazard mitigation of seismically excited buildings equipped with a hybrid control system. The wavelet-based fuzzy neuro controller is developed through the integration of discrete wavelet transform, artificial neural network, and a Takagi-Sugeno fuzzy controller. The hybrid control system is an integrated model of an actuator, a tuned mass damper, and viscous liquid dampers: an active tuned mass damper (ATMD) is located on the top floor of the structure and viscous liquid dampers are located on each floor. To demonstrate the effectiveness of the proposed wavelet-based adaptive neuro-fuzzy inference system (WANFIS) controller, an eight-story building employing passive viscous liquid dampers as well as an ATMD is investigated. A variety of earthquakes such as an artificial earthquake, the 1940 El-Centro, Kobe, Northridge, and Hachinohe earthquakes are used as disturbance signals. It is demonstrated that the WANFIS controller is effective in reducing the structural responses of the hybrid structure system subjected to a variety of disturbances. [ABSTRACT FROM AUTHOR]

This paper presents a new inversion approach to estimate the depth profile of a fatigue crack from the output signal of an alternating current field measurements (ACFM) probe. We first show that the mapping between the ACFM signals and corresponding crack depth profiles can be expressed by an affine transformation, then we utilize a generalized version of the fuzzy alignment algorithm (FAA) to find this mapping more efficiently. The proposed method is then extended using the Dave's objective function previously proposed for the fuzzy clustering. To evaluate the accuracy and robustness of the proposed inversion technique we reconstruct cracks of arbitrary shapes from both simulated and measured ACFM signals. The results show that the proposed approach is more robust and accurate than existing methods and gives reasonable results even in the absence of sufficient training data. [ABSTRACT FROM PUBLISHER]

The objective of this paper is to review some important progress with H infinity control methods and to derive an artificial algorithm for dealing with stability problems of multiple time-delay structural systems. Neural network modeling and linear deferential inclusion are utilized to represent a multiple time-delay structural system. We can derive a novel stability criterion using linear matrix inequality methods. Moreover, the H∞ control performance is ensured for multiple time-delay structural systems based on traditional Lyapunov theory. [ABSTRACT FROM PUBLISHER]

This paper presents an application of artificial neural network for the classification of single channel EMG signal in the context of hand motion detection. Seven statistical input features that are extracted from the preprocessed single channel EMG signals recorded for four predefined hand motions have been used for neural network classifier. Different structures of neural network, based on the number of hidden neurons and two prominent training algorithms, have been considered in the research to find out their applicability for EMG signal classification. The classification performances are analyzed for different architectures of neural network by considering the number of input features, number of hidden neurons, learning algorithms, correlation between network outputs and targets, and mean square error. Between the Levenberg-Marquardt and scaled conjugate gradient learning algorithms, the aforesaid algorithm shows better classification performance. The outcomes of the research show that the optimal design of Levenberg-Marquardt based neural network classifier can perform well with an average classification success rate of 88.4%. A comparison of results has also been presented to validate the effectiveness of the designed neural network classifier to discriminate EMG signals. [ABSTRACT FROM AUTHOR]

Abstract: It is known that most of the key problems in visual servo control of robots are related to the performance analysis of the system considering measurement and modeling errors. In this paper, the development and performance evaluation of a novel intelligent visual servo controller for a robot manipulator using neural network Reinforcement Learning is presented. By implementing machine learning techniques into the vision based control scheme, the robot is enabled to improve its performance online and to adapt to the changing conditions in the environment. Two different temporal difference algorithms (Q-learning and SARSA) coupled with neural networks are developed and tested through different visual control scenarios. A database of representative learning samples is employed so as to speed up the convergence of the neural network and real-time learning of robot behavior. Moreover, the visual servoing task is divided into two steps in order to ensure the visibility of the features: in the first step centering behavior of the robot is conducted using neural network Reinforcement Learning controller, while the second step involves switching control between the traditional Image Based Visual Servoing and the neural network Reinforcement Learning for enabling approaching behavior of the manipulator. The correction in robot motion is achieved with the definition of the areas of interest for the image features independently in both control steps. Various simulations are developed in order to present the robustness of the developed system regarding calibration error, modeling error, and image noise. In addition, a comparison with the traditional Image Based Visual Servoing is presented. Real world experiments on a robot manipulator with the low cost vision system demonstrate the effectiveness of the proposed approach. [Copyright &y& Elsevier]

Abstract: Feature selection and classification techniques have been studied independently without considering the interaction between both procedures, which leads to a degraded performance. In this paper, we present a new neural network approach, which is called an algorithm learning based neural network (ALBNN), to improve classification accuracy by integrating feature selection and classification procedures. In general, a knowledge-based artificial neural network operates on prior knowledge from domain experience, which provides it with better starting points for the target function and leads to better classification accuracy. However, prior knowledge is usually difficult to identify. Instead of using unknown background resources, the proposed method utilizes prior knowledge that is mathematically calculated from the properties of other learning algorithms such as PCA, LARS, C4.5, and SVM. We employ the extreme learning machine in this study to help obtain better initial points faster and avoid irrelevant time-consuming work, such as determining architecture and manual tuning. ALBNN correctly approximates a target hypothesis by both considering the interaction between two procedures and minimizing individual procedure errors. The approach produces new relevant features and improves the classification accuracy. Experimental results exhibit improved performance in various classification problems. ALBNN can be applied to various fields requiring high classification accuracy. [Copyright &y& Elsevier]

Abstract: This study aims to develop a PD control system based on the neural network algorithm to reduce the ship rolling motion in the desired track through rudder and fin actions. The mathematical model including sea-keeping and maneuvering characteristics is developed in the present paper and the nonlinear time history of ship motions is solved by the fourth order Runge–Kutta method. In order to achieve the purpose of roll reduction and track keeping, the rudder and fin stabilizer are used as the control tools for the ship advancing in the seaway. In addition, the PD controller based on the self-tuning neural network algorithm is applied to achieve the goals of multi-input and multi-output in the control system. Four different types of control modes on a container ship model are studied and the performances are investigated in different sea states. The results indicate that the PD controller based on the self-tuning neural network algorithm applying to the stabilizer fin control for roll reduction and the rudder control for track keeping in the seaway would be suggested. [Copyright &y& Elsevier]

In this paper, the method of the graphical interpretation of the single-layer network weights is introduced. It is shown that the network parameters can be converted to the image and their particular elements are the pixels. For this purpose, weight-to-pixel conversion formula is used. Moreover, new weights' modification method is proposed. The weight coefficients are computed on the basis of pixel values for which image filtration algorithms are implemented. The approach is applied to the weights of three types of the models: single-layer network, two-layer backpropagation network and the hybrid network. The performance of the models is then compared on two independent data sets. By means of the experiments, it is presented that the adjustment of the weights to new values decreases test error value compared to the error obtained for initial set of weights. [ABSTRACT FROM AUTHOR]

Abstract: The Gravitational Search Algorithm (GSA) is a novel heuristic optimization method based on the law of gravity and mass interactions. It has been proven that this algorithm has good ability to search for the global optimum, but it suffers from slow searching speed in the last iterations. This work proposes a hybrid of Particle Swarm Optimization (PSO) and GSA to resolve the aforementioned problem. In this paper, GSA and PSOGSA are employed as new training methods for Feedforward Neural Networks (FNNs) in order to investigate the efficiencies of these algorithms in reducing the problems of trapping in local minima and the slow convergence rate of current evolutionary learning algorithms. The results are compared with a standard PSO-based learning algorithm for FNNs. The resulting accuracy of FNNs trained with PSO, GSA, and PSOGSA is also investigated. The experimental results show that PSOGSA outperforms both PSO and GSA for training FNNs in terms of converging speed and avoiding local minima. It is also proven that an FNN trained with PSOGSA has better accuracy than one trained with GSA. [Copyright &y& Elsevier]

Along with the development of monitoring technologies, numerous measured data pour into monitoring system and form the high-volume and open-ended data stream. Usually, abnormal condition of monitored system can be characterized by the density variation of measured data stream. However, traditional density estimation methods can not dynamically track density variation of data stream due to the limitation of processing time and computation memory. In this paper, we propose a new density estimation neural network to continuously estimate the density of streaming data in a time-based sliding window. The network has a feedforward structure composed of discretization, input and summation layer. In the discretization layer, value range of data stream is discretized to network nodes with equal intervals. Measured data in the predefined time window are pushed into input layer and updated with the window sliding. In summation layer, the activation results between input neurons and discretization neurons are summed up and multiplied by a weight factor. The network outputs the kernel density estimators of sliding segment in data stream and achieves a one-pass estimation algorithm consuming constant computation memory. By subnet separation and local activation, computation load of the network is significantly reduced to catch up the pace of data stream. The nonlinear statistics, quantile and entropy, which can be consecutively figured out with the density estimators output by the density estimation neural network, are calculated as condition indictors to track the density variation of data stream. The proposed method is evaluated by a simulated data stream consisting of two mixing distribution data sets and a pressure data stream measured from a centrifugal compressor respectively. Results show that the underlying anomalies are successfully detected. [ABSTRACT FROM AUTHOR]

Abstract: Two new on-line algorithms for adaptive principal component analysis (APCA) are proposed and discussed in order to solve the problem of on-line industrial process monitoring in this paper. Both the algorithms have the capability of extracting principal component eigenvectors on-line in a fixed size sliding data window with high dimensional input data. The first algorithm is based on the steepest gradient descent approach, which updates the covariance matrix with deflation transformation and on-line iteration. Based on neural networks, the second algorithm constructs the input data sequence with an on-line iteration method and trains the neural network in every data frame. The convergence of the two algorithms is then analyzed and the simulations are given to illustrate the effectiveness of the two algorithms. At last, the applications of the two algorithms are discussed. [Copyright &y& Elsevier]

Abstract: Ensemble approaches to classification have attracted a great deal of interest recently. This paper presents a novel method for designing the neural network ensemble using coevolutionary algorithm. The bootstrap resampling procedure is employed to obtain different training subsets that are used to estimate different component networks of the ensemble. Then the cooperative coevolutionary algorithm is developed to optimize the ensemble model via the divide-and-cooperative mechanism. All component networks are coevolved in parallel in the scheme of interacting co-adapted subpopulations. The fitness of an individual from a particular subpopulation is assessed by associating it with the representatives from other subpopulations. In order to promote the cooperation of all component networks, the proposed method considers both the accuracy and the diversity among the component networks that are evaluated using the multi-objective Pareto optimality measure. A hybrid output-combination method is designed to determine the final ensemble output. Experimental results illustrate that the proposed method is able to obtain neural network ensemble models with better classification accuracy in comparison with currently popular ensemble algorithms. [Copyright &y& Elsevier]

The dressing process is affected on the wheel surface topography and the selection of optimal dressing parameters makes increasing in the grinding performance. In this paper, a feed forward backpropagation neural network (FFBP-NN) and a simulated annealing (SA) algorithm were used for simultaneously minimize the tangential grinding force as well as minimize the surface roughness. First, the FFBP-NN was developed using the data generated based on experimental observations. In the experiments procedure, the grinding conditions were constant and only the disk dressing conditions were varied. Then, a SA algorithm was applied to the FFBP-NN for solving the optimal dressing parameters. In order to demonstrate the model performance, several initial trial values have tested. Results show that the proposed model has an acceptable performance to optimize the grinding process. [ABSTRACT FROM AUTHOR]

Active queue control aims to improve the overall communication network throughput, while providing lower delay and small packet loss rate. The basic idea is to actively trigger packet dropping (or marking provided by explicit congestion notification (ECN)) before buffer overflow. In this paper, two artificial neural networks (ANN)-based control schemes are proposed for adaptive queue control in TCP communication networks. The structure of these controllers is optimized using genetic algorithm (GA) and the output weights of ANNs are optimized using particle swarm optimization (PSO) algorithm. The controllers are radial bias function (RBF)-based, but to improve the robustness of RBF controller, an error-integral term is added to RBF equation in the second scheme. Experimental results show that GA- PSO-optimized improved RBF (I-RBF) model controls network congestion effectively in terms of link utilization with a low packet loss rate and outperforms Drop Tail, proportional-integral (PI), random exponential marking (REM), and adaptive random early detection (ARED) controllers. [ABSTRACT FROM AUTHOR]

Abstract: Increasing demand of bandwidth in communication satellites has forced satellite links to be designed in Ku bands and above. But at these frequencies, rain and other tropospheric elements result in large attenuation. To mitigate the tropospheric attenuation of microwave satellite signals above 10GHz using any standard Fade Mitigation Technique (FMT), it is essential to have a priori knowledge about the level of attenuation. Hence, short-term rain attenuation prediction models play a key role in maintaining the link in which necessary compensation can be applied depending on the early information of attenuation. This paper presents a method of attenuation prediction using Adaptive Artificial Neural Network. Here In situ Learning Algorithm (ILA) has been used to enable the system to track the non-stationary nature of the attenuation. To validate this, Ku Band data, collected at three different sites in India have been used for the purpose of prediction. The performance of the algorithm is determined through the estimation of prediction accuracy by comparing the predicted values with the measured data. Results obtained using the mentioned technique shows considerably good accuracy even up to 20s of prediction interval with acceptable ratio between the under and over predictions. The prediction performance is evaluated for different prediction intervals. Furthermore the present model is also compared with the persistence model and the relative performance is quantified. [Copyright &y& Elsevier]

A new method for real-time prediction of uncertain network transmission time delays and a method for closed-loop control of manufacturing and industrial plants through networks are introduced. The proposed delay prediction method is based on the multilayer perceptron neural model. In order to minimize the number of neurons in the first layer of the network and hence reducing the computational burden in a real-time implementation, a method for determination of the Markov order of the time delay sequence is presented. Using the predicted delay, and a zero-order hold equivalent discrete-time model of the plant, a time varying state feedback control algorithm with a real-time gain updating strategy is proposed. A sufficient condition for closed-loop stability is also derived using the switching theorem for linear systems. The proposed method is shown, through two industrial networked case studies, namely, a DC motor driving a transportation roller for paper sheets and a milling machine. Simulation studies depict the efficacy of the proposed method in controlling such challenging problems. [ABSTRACT FROM AUTHOR]

Abstract: In mammography diagnosis systems, high False Negative Rate (FNR) has always been a significant problem since a false negative answer may lead to a patient''s death. This paper is directed towards the development of a novel Computer-aided Diagnosis (CADx) system for the diagnosis of breast masses. It aims at intensifying the performance of CADx algorithms as well as reducing the FNR by utilizing Zernike moments as descriptors of shape and margin characteristics. The input Regions of Interest (ROIs) are segmented manually and further subjected to a number of preprocessing stages. The outcomes of preprocessing stage are two processed images containing co-scaled translated masses. Besides, one of these images represents the shape characteristics of the mass, while the other describes the margin characteristics. Two groups of Zernike moments have been extracted from the preprocessed images and applied to the feature selection stage. Each group includes 32 moments with different orders and iterations. Considering the performance of the overall CADx system, the most effective moments have been chosen and applied to a Multi-layer Perceptron (MLP) classifier, employing both generic Back Propagation (BP) and Opposition-based Learning (OBL) algorithms. The Receiver Operational Characteristics (ROC) curve and the performance of resulting CADx systems are analyzed for each group of features. The designed systems yield Az=0.976, representing fair sensitivity, and Az=0.975 demonstrating fair specificity. The best achieved FNR and FPR are 0.0% and 5.5%, respectively. [Copyright &y& Elsevier]

Multiple algorithms have been developed to correct user's typing mistakes. However, an optimal solution is hardly identified among them. Moreover, these solutions rarely produce a single answer or share common results, and the answers may change with time and context. These motivated this research to synthesize some distinct word correction algorithms to produce an optimal prediction based on database updates and neural network learning. In this paper, three distinct typing correction algorithms are integrated as a pilot research with key factors such as Time Change, Context Change and User Feedback being considered. Experimental results show that the developed WLR model (i.e., word-list neural network ranking model) achieves the best results in comparison with Levenshtein distance, Metaphone. and Two-Gram prediction algorithms throughout all stages. It achieves 57.50% Ranking First Hitting Rate with samples of category one and a best Ranking First Hitting Rate of 74.69% within category four. Further work is recommended to extend the number of parameters and integrate more algorithms to achieve a higher level of accuracy. [ABSTRACT FROM AUTHOR]

Abstract: Nowadays, pavement distresses classification becomes more important, as the computational power increases. Recently, multi-resolution analysis such as wavelet decompositions provides very good multi-resolution analytical tools for different scales of pavement analysis and distresses classification. In this paper an expert system is proposed for pavement distress classification. A radon neural network, based on wavelet transform expert system is used for increasing the effectiveness of the scale invariant feature extraction algorithm. Wavelet modulus is calculated and Radon transform is then applied to the wavelet modulus. The features and parameters of the peaks are finally used for training and testing the neural network. Experimental results demonstrate that the proposed expert system is an effective method for pavement distress classification. The test performances of this study show the advantages of proposed expert system: it is rapid, easy to operate, and have simple structure. [Copyright &y& Elsevier]

Abstract: Understanding how the various factors in a supply chain contribute to the overall performance of its operation has become an important topic in management science research nowadays. In this paper, we propose and apply a two-stage methodology to an industrial survey data set to investigate relations among the key factors in a supply chain model. Precisely, we use the PC-algorithm to discover the connectivity relation among the factors of interest in the supply chain model. Critical factors in the model are then identified, and we then utilize the neural network to quantify the relative importance of some of the factors in predicting the critical factors. An advantage of our proposed method is that it frees up the researcher from making subjective decisions in his or her analysis, for example, from the needs of specifying plausible initial path models required in a structural equation modeling analysis (which is usually used in business and management research) and of selecting factors for the subsequent predictive modeling. We envision that the analysis results can aid a decision maker in optimizing the system performance by suggesting to the decision maker which ones of the factors are the important ones that he or she should devote more resources and efforts on. [Copyright &y& Elsevier]

Abstract: Uncertain ground conditions represent the primary source of risk in underground tunnel construction. However, this problem can be solved by developing an accurate, probabilistic description of the geology. This paper presents a general model for probability based determination of tunnel geology that can be used as a basis for developing more effective decision support systems for tunneling design and construction. The proposed model is based on a Hidden Markov Model (HMM) and a neural network (NN). An approximate inference technique – a Particle Filter (PF) Algorithm – is used to simulate the geological parameters. This model overcomes the deficiencies of existing models by readily incorporating all available geologic information and updating geologic predictions based on observations given by the neural network. In order to validate the proposed model, the “Drainage Water Tunnel Project” at Zhong-He, Taipei, Taiwan was used. The results showed that the Neural-HMM model provides high accuracy in geological prediction. [Copyright &y& Elsevier]

Abstract: This paper investigates failure in Carbon Fibre Reinforced Plastics CFRP using Acoustic Emission (AE). Signals have been collected and post-processed for various test configurations: tension, Compact Tension (CT), Compact Compression (CC), Double Cantilever Beam (DCB) and four-point bend End Notched Flexure (4-ENF). The signals are analysed with three different pattern recognition algorithms: k-means, Self Organising Map (SOM) combined with k-means and Competitive Neural Network (CNN). The SOM combined with k-means appears as the most effective of the three algorithms. The results from the clustering analysis follow patterns found in the peak frequencies distribution. A detailed study of the frequency content of each test is then performed and the classification of several failure modes is achieved. [Copyright &y& Elsevier]

In this paper, a neural network is trained and validated using a low end and inexpensive microcontroller. The well-known backpropagation algorithm is implemented to train a neural network model. Both the training and the validation parts are shown through an alphanumeric liquid crystal display. A chemical process was chosen as a realistic nonlinear system to demonstrate the feasibility, and the performance of the results found using the microcontroller. A comparison was made between the microcontroller and the computer results. [ABSTRACT FROM AUTHOR]

In this paper, a novel multiscale algorithm to simulate accumulation of trabecular bone crack density and crack length at macroscopic scale during cyclic loading is developed. The method is based on finite element analysis and neural network computation to link mesoscopic (trabecular level) and macroscopic (whole femur) scales. The finite element calculation is performed at the macroscopic level and a trained neural network incorporated into the finite element code Abaqus is employed as a numerical device to perform the local mesoscopic computation. Based on a set of mesoscale simulations of representative volume elements obtained by digital image-based modeling technique using µ-CT and voxel finite element, a neural network is trained to approximate the local finite element responses. The input data for the artificial neural network are the applied stress, the stress orientation and the cycle frequency. The output data are the averaged crack density and crack length at a given site of the bone. Copyright © 2010 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Abstract: In this paper, the convergence of a new back-propagation algorithm with adaptive momentum is analyzed when it is used for training feedforward neural networks with a hidden layer. A convergence theorem is presented and sufficient conditions are offered to guarantee both weak and strong convergence result. Compared with existing results, our convergence result is of deterministic nature and we do not require the error function to be quadratic or uniformly convex. [ABSTRACT FROM AUTHOR]

This paper describes an arrangement of a one-dimensional distributive tactile sensing system that can be used to determine an applied position of a point load of a constant magnitude. The performance of the system was examined using inputs derived from a mathematical model and a back propagation neural network as an interpretation algorithm. Performances of the system with 2-8 inputs with and without an application of principal component analysis (PCA) as a preprocessor were examined. For each number of inputs, four sets of sensing positions were explored and the accuracies in determining an applied load position were compared. It was found that the system was able to determine an applied load position with errors in the range of 1.0-3.1 mm depending on the number of inputs and the method of inputting data. The error decreased with an increase in the number of inputs. It was found that input preprocessing by PCA impaired the performance. Systematically chosen and optimal sets of sensing positions resulted in the most desirable performance and their performances were compa- rable. Amongst the sets of input positions explored, random positions yielded the highest errors. Random positions also resulted in the largest difference between the first two principal components. [ABSTRACT FROM AUTHOR]

general neural network based methodology that has the potential to replace the computationally-intensive site-specific seismic analysis of structures is proposed in this paper. The basic framework of the methodology consists of a feed forward back propagation neural network algorithm with one hidden layer to represent the seismic potential of a region and soil amplification effects. The methodology is implemented and verified with parameters corresponding to Delhi city in India. For this purpose, strong ground motions are generated at bedrock level for a chosen site in Delhi due to earthquakes considered to originate from the central seismic gap of the Himalayan belt using necessary geological as well as geotechnical data. Surface level ground motions and corresponding site-specific response spectra are obtained by using a one-dimensional equivalent linear wave propagation model. Spectral acceleration values are considered as a target parameter to verify the performance of the methodology. Numerical studies carried out to validate the proposed methodology show that the errors in predicted spectral acceleration values are within acceptable limits for design purposes. The methodology is general in the sense that it can be applied to other seismically vulnerable regions and also can be updated by including more parameters depending on the state-of-the-art in the subject. [ABSTRACT FROM AUTHOR]

Abstract: Minor subspace analysis (MSA) is a statistical method for extracting the subspace spanned by all the eigenvectors associated with the minor eigenvalues of the autocorrelation matrix of a high-dimension vector sequence. In this paper, we propose a self-stabilizing neural network learning algorithm for tracking minor subspace in high-dimension data stream. Dynamics of the proposed algorithm are analyzed via a corresponding deterministic continuous time (DCT) system and stochastic discrete time (SDT) system methods. The proposed algorithm provides an efficient online learning for tracking the MS and can track an orthonormal basis of the MS. Computer simulations are carried out to confirm the theoretical results. [Copyright &y& Elsevier]

The process of pressed assembly design is in a causal relation with the evaluation of friction coefficient. The evaluation of this parameter can be carried out by using various methods. In this paper, the algorithm for friction coefficient determination is analyzed. Analyses are based on roughness and hardness of tactile surfaces, with the implementation of a hybrid system based on neural networks. The value of this parameter is usually determined according to literature data. Many authors propose different values for this coefficient. In this case, the ratio between the minimum value and the maximum value is 1:17, and it depends on the designer’s decision. The implementation of neural networks results in the determination of friction coefficient and error in the range of 20%. [ABSTRACT FROM AUTHOR]

The larger display devices, the more noticeable artifacts such as false contours, block artifacts, and other types of noises. This paper proposes a false contour reduction algorithm using neural networks (NNs) and adaptive bidirectional smoothing. The proposed algorithm consists of two parts: false contour detection and reduction parts. In the false contour detection part, false contour candidate pixels are detected using the directional contrast features. The false contour reduction part is composed of two steps: NN processing and bi-directional filtering. In the first step, false contours are reduced by pixelwise processing using NNs. In the second step, bi-directional smoothing is applied to a neighboring region of the false contour. Computer simulations with several test images show the effectiveness of the proposed false contour reduction algorithm in terms of the visual quality of result images, edge maps detected by Sobel masks, the peak signal-to-noise ratio, the structural similarity, and the computation time. [ABSTRACT FROM AUTHOR]