Search

Oil-immersed paper is a weak link in internal insulation of oil immersed power transformer, whose aging state distributes inhomogeneously. However, most of the current assessment methods fail to consider the spatial distribution of aging areas inside transformer. Aiming at this problem, this paper proposes the inversion detection method to obtain the resistivity of oil-immersed paper in various regions nondestructively, and uses the resistivity of oil-immersed paper to assess directly the insulation state of oil immersed power transformers. Finite element method is applied to establish the mapping relationship between the resistivity of oil-immersed paper in different regions, and the dielectric loss factor obtained from the port of transformer. Back propagation neural network is used to learn inversely this mapping relationship so as to calculate the resistivity. Applying the proposed method on a 10-kV transformer, the calculated results are close to the measured values. The study shows that the inversion method can effectively calculate the oil-immersed resistivity in various regions, which represent the aging state in different regions of the transformer. [ABSTRACT FROM AUTHOR]

Oil-paper insulation ageing of transformer has characteristics of spatial locality due to uneven temperature distribution. The transformer is sealed, it is difficult to obtain the local ageing state of oil-paper insulation by sampling directly. Currently, there are no effective detection method to detect the insulation state of local regions. The oil-immersed paper resistivity inversion method can nondestructively assess the local state of oil-paper insulation. However, the current inversion methods still have shortcomings and need to be improved. In this paper, combining the back propagation neural network (BPNN) and Newton downhill method (NDM), an improved inversion method is proposed to optimize the current inversion method, so as to reduce the effect of initial value on it and make calculated results converge better. Finally, a 3D 10 kV transformer model is established to verify the effectiveness of the proposed method. The resistivity calculated by the improved method show that the method can provide initial resistivity values for technicians and improve the convergence and accuracy of current inversion method, which is more suitable for extending to the engineering practice. [ABSTRACT FROM AUTHOR]

Time-of-use (TOU) is an effective price-based demand response strategy. A reasonable design of TOU strategy can effectively reduce the peak-valley difference, and then produce a lot of benefits (such as delaying power grid investment, reducing interruption cost, and improving reliability). However, changing peak-valley period has a great influence on the peak-valley difference and power supply reliability of power system. Therefore, this paper aims to investigate the optimal modification of peak-valley period considering reliability loss under multiple TOU schemes. Firstly, this paper presents a clustering model and algorithm of optimal load curve based on a minimum error iteration method. Secondly, an optimal modification of peak-valley period based on a dynamic load point method is proposed, and the traditional peak-valley difference is replaced by the global peak-valley difference to calculate the objective function. Thirdly, this paper establishes a load–reliability relation fitting model based on the back propagation neural network. Finally, the effectiveness and correctness of the proposed method are investigated by the Roy Billinton test system and reliability test system. [ABSTRACT FROM AUTHOR]

In a relational database, tuples are called “duplicate” if they describe the same real-world entity. If such duplicate tuples are observed, it is recommended to remove them and to replace them with one tuple that represents the joint information of the duplicate tuples to a maximal extent. This remove-and-replace operation is called a fusion operation. Within the setting of a relational database management system, the removal of the original duplicate tuples can breach referential integrity. In this paper, a strategy is proposed to maintain referential integrity in a semantically correct manner, thereby optimizing the quality of relationships in the database. An algorithm is proposed that is able to propagate a fusion operation through the entire database. The algorithm is based on a framework of first and second order fusion functions on the one hand, and conflict resolution strategies on the other hand. It is shown how classical strategies for maintaining referential integrity, such as delete cascading, are highly specialized cases of the proposed framework. Experimental results are reported that (i) show the efficiency of the proposed algorithm and (ii) show the differences in quality between several second order fusion functions. It is shown that some strategies easily outperform delete cascading. [ABSTRACT FROM PUBLISHER]

In underwater acoustic (UWA) adaptive communication system, due to time-varying channel, the transmitter often has outdated channel state information (CSI), which results in low efficiency. UWA channels are much more difficult to estimate and predict than terrestrial wireless channels, given the more severe multipath environments with varying propagation speeds in different locations, non-linear propagation paths, several-order higher propagation latency, mobile transceiver and obstacles in the sea, etc. To handle the complexity, this paper proposes an efficient online CSI prediction model for UWA CSI prediction considering the complicated correlationship of UWA channels in both the time and frequency domains. This paper designs a learning model called CsiPreNet, which is an integration of a one-dimensional convolutional neural network (CNN) and a long short term memory (LSTM) network. The performance is compared with the widely used recursive least squares (RLS) predictor, the approximate linear dependency recursive kernel least-squares (ALD-KRLS), and two common conventional deep neural networks (DNN) predictors, i.e., back propagation neural network (BPNN) and LSTM network using the measured data recorded in the South China Sea. To validate the efficacy of prediction, we investigate the prediction of CSI in simulated downlink UWA orthogonal frequency division multiple access (OFDMA) systems. Specifically, the measured UWA channel is used in the OFDMA system. A joint subcarrier-bit-power adaptive allocation scheme is used for resource allocation. To further improve the performance, we develop an offline-online prediction scheme, enabling the prediction results to be more stable. Simulation and experimental results show that the CsiPreNet has superior performance than the existing solutions, thanks to its capability in capturing both the temporal and frequency correlation of the UWA CSIs. [ABSTRACT FROM AUTHOR]

It is a great challenge to differentiate partial discharge (PD) induced by different types of insulation defects in high-voltage cables. Some types of PD signals have very similar characteristics and are specifically difficult to differentiate, even for the most experienced specialists. To overcome the challenge, a convolutional neural network (CNN)-based deep learning methodology for PD pattern recognition is presented in this paper. First, PD testing for five types of artificial defects in ethylene-propylene-rubber cables is carried out in high voltage laboratory to generate signals containing PD data. Second, 3500 sets of PD transient pulses are extracted, and then 33 kinds of PD features are established. The third stage applies a CNN to the data; typical CNN architecture and the key factors which affect the CNN-based pattern recognition accuracy are described. Factors discussed include the number of the network layers, convolutional kernel size, activation function, and pooling method. This paper presents a flowchart of the CNN-based PD pattern recognition method and an evaluation with 3500 sets of PD samples. Finally, the CNN-based pattern recognition results are shown and the proposed method is compared with two more traditional analysis methods, i.e., support vector machine (SVM) and back propagation neural network (BPNN). The results show that the proposed CNN method has higher pattern recognition accuracy than SVM and BPNN, and that the novel method is especially effective for PD type recognition in cases of signals of high similarity, which is applicable for industrial applications. [ABSTRACT FROM AUTHOR]

This paper is devoted to solving a full-wave inverse scattering problem (ISP), which is aimed at retrieving permittivities of dielectric scatterers from the knowledge of measured scattering data. ISPs are highly nonlinear due to multiple scattering, and iterative algorithms with regularizations are often used to solve such problems. However, they are associated with heavy computational cost, and consequently, they are often time-consuming. This paper proposes the convolutional neural network (CNN) technique to solve full-wave ISPs. We introduce and compare three training schemes based on U-Net CNN, including direct inversion, backpropagation, and dominant current schemes (DCS). Several representative tests are carried out, including both synthetic and experimental data, to evaluate the performances of the proposed methods. It is demonstrated that the proposed DCS outperforms the other two schemes in terms of accuracy and is able to solve typical ISPs quickly within 1 s. The proposed deep-learning inversion scheme is promising in providing quantitative images in real time. [ABSTRACT FROM AUTHOR]

In this paper, we propose a microwave imaging technique that involves an adaptive back-projection (BP) algorithm based on a climb method. In general, the range migration algorithm (RMA) provides high computational efficiency and high accuracy for synthetic aperture radar (SAR) imaging. However, RMA involves wavenumber domain interpolation, which may adversely affect the accuracy of the target imaging. The BP algorithm, which has its origins in medical imaging techniques, has been studied to overcome this problem in SAR imaging, because it avoids interpolation associated with domain conversion, but it also has a huge computational cost. Therefore, this paper proposes adaptive processing based on a climb method to find the target indicator as real target responses. The real target responses are predetermined and the raw data are reproduced to reduce the computational cost and to improve the image quality. Finally, the superiority of proposed algorithm is demonstrated using both the simulated and experimental data. [ABSTRACT FROM AUTHOR]

Cooperative localization is an important technique in wireless networks. However, there are always errors in network node localization, which will spatially propagate among network nodes when performing network localization. In this paper, we study the spatial error propagation (EP) characteristics of network localization in terms of Fisher information. First, the spatial propagation function is proposed to reveal the spatial cooperation principle of network localization. Second, the convergence property of spatial localization information propagation (SLIP) is analyzed to shed light on the performance limits of network localization through spatial information propagation. It is shown that 1) the network localization error propagates in the way of Ohm's law in electric circuit theory, where the measurement accuracy, node location accuracy, and geometric-resolution information behave like the resistances connected in parallel or series; 2) the network location error gradually diminishes with spatial localization cooperation procedures, due to the cooperative localization information propagation; and 3) the essence of spatial localization cooperation among network nodes is the spatial propagation of localization information. [ABSTRACT FROM PUBLISHER]

The 17 papers in this special section examine wireless power transfer (WPT)for smart industrial and home applications with emphasis on e theoretical analysis, design, control strategies, management, coordination, and communication of WPT systems for smart industrial and household applications. [ABSTRACT FROM AUTHOR]

In this paper, we have proposed a spin-orbit torque driven skyrmionic device that acts as a synaptic element in a hardware artificial neural network (ANN) and compared its performance with similar domain wall-based device in the presence of defects in the material that can pin the domain wall or the skyrmion. Next, we have simulated a feedforward ANN with either a domain wall device or a skyrmionic device as the synaptic element and solved a standard digit recognition problem with it using the back-propagation algorithm to show that a skyrmionic synapse-based ANN consumes two orders of magnitude lower energy than a domain wall synapse-based ANN. [ABSTRACT FROM AUTHOR]

This paper proposes a nonlinear flux linkage model in 2-D plane for the planar switched reluctance motor (PSRM). The inputs of the proposed model are the 2-D positions and the current, and the output is the flux linkage. The proposed model is established via a cascade-forward backpropagation neural network (CFNN). The designed CFNN consists of four layers: one input layer, two hidden layers, and one output layer. The first hidden layer has 20 neurons with a tan-sigmoid transfer function, and the second hidden layer has 20 neurons with a log-sigmoid transfer function. The output layer is a pure linear layer. The sample set with 179 755 samples is obtained experimentally in a dSPACE-based PSRM system by applying the dc excitation method. The sample set is divided into three sets. 35% and 30% of the samples are randomly chosen as the training sample set and validation sample set, respectively, and the remaining samples are utilized as the test sample set to assess the generalization performance of the CFNN-based model. According to the results of the test sample set, the maximum relative error is 11.05% and the mean relative error is 0.42% when the current ranges from 1 to 9 A. The CFNN has the capability to build a multi-input nonlinear model. The CFNN-based model is capable of reflecting the variations of flux linkage in 2-D plane caused by manufacturing tolerances. The effectiveness of the CFNN-based model is finally verified. [ABSTRACT FROM AUTHOR]

The reconstruction of a source embedded within a multipath environment, which is created by inserting a grid of point scatterers in the scene, is addressed. In particular, the source Fourier spectrum is assumed known so that the focus here is on the reconstruction of the spatial support. As well documented, multipath can allow for resolution improvement. However, it also gives rise to artifacts when a backpropagation-like imaging is adopted. In this paper, we study in detail how resolution improvement and artifacts depend on the grid layout by employing a weighted backpropagation algorithm. More in detail, stationary phase arguments are used to predict the reconstruction leading order terms to which resolution improvement is linked. Moreover, it is shown that artifacts are mainly due to high-order terms and are dependent on the point scatterers’ arrangement. The nature of such artifacts is studied, and a simple way to mitigate their role (without resolution loss) is introduced; it consists in a suitable nonuniform grid arrangement with a “hole in the center.” Backpropagation is then compared with an inverse filtering imaging based on the truncated singular value decomposition (TSVD) of the radiation operator. It is shown that the TSVD is less prone to artifacts and can, in principle, allow for a higher resolution improvement. However, when model error (due to multiple scattering between the elements of the grid) and/or noise corrupt data, backpropagation performs definitely better. The theoretical findings are supported by an extensive numerical analysis. In particular, to keep the figures simple, we consider only 2-D cases. [ABSTRACT FROM AUTHOR]

Sparsity-constrained optimization problems are common in machine learning, such as sparse coding, low-rank minimization and compressive sensing. However, most of previous studies focused on constructing various hand-crafted sparse regularizers, while little work was devoted to learning adaptive sparse regularizers from given input data for specific tasks. In this paper, we propose a deep sparse regularizer learning model that learns data-driven sparse regularizers adaptively. Via the proximal gradient algorithm, we find that the sparse regularizer learning is equivalent to learning a parameterized activation function. This encourages us to learn sparse regularizers in the deep learning framework. Therefore, we build a neural network composed of multiple blocks, each being differentiable and reusable. All blocks contain learnable piecewise linear activation functions which correspond to the sparse regularizer to be learned. Furthermore, the proposed model is trained with back propagation, and all parameters in this model are learned end-to-end. We apply our framework to multi-view clustering and semi-supervised classification tasks to learn a latent compact representation. Experimental results demonstrate the superiority of the proposed framework over state-of-the-art multi-view learning models. [ABSTRACT FROM AUTHOR]

Dealing with radar backscattering from trees, the Wong model is a mixing of Gaussian spectra with parameters deduced from considerations on motions of branches and leaves. Very detailed experiments by Narayanan show gaps with this model. We show that autocorrelation functions by Narayanan are very well fitted by functions in the form \exp[-\vert\tau/\tau0\vert^{\alpha}], 0 < \alpha \leq 2. In this paper, we prove that the random propagation time theory explains this property. I have shown in other papers that this theory is available to study power spectra in acoustics, ultrasonics, and electromagnetics. [ABSTRACT FROM PUBLISHER]

The aim of this paper is to investigate the estimation of the variation of ground resistance throughout the year by using artificial neural networks (ANNs). An ANN was trained, validated, and tested with different training algorithms by using experimental data of soil resistivity, ground resistance, and rainfall in order to select the optimum training algorithm and the respective parameters and predict the behavior of the ground resistance of a single rod. Moreover, a sensitivity analysis of the proposed ANN was carried out in order to determine the impact of certain factors on the efficiency of the ANN. The high value of the correlation index between estimated and experimental values demonstrates the high efficiency of the ANN. The proposed methodology based on ANN is a useful tool for the estimation of the grounding resistance during the year in case of difficulties in measuring its value. [ABSTRACT FROM AUTHOR]

Massive multiple-input multiple-output (M-MIMO) aided positioning technologies have been recently recognized as promising solutions to fulfil the high accuracy requirement of indoor localization systems. In this paper, exploiting a distributed M-MIMO framework, we propose to employ a deep belief network (DBN) to analyze the received signal strengths (RSS) generated by a diffraction model, where the impact from interfering persons on the targeted user equipment (UE) is considered. Next, the preliminary DBN estimates are forwarded to a long-short term memory network (LSTMN), where the trajectory information of the targeted UE can be extracted based on much less historical trajectory information than existing solutions. Then, the three-dimension (3D) coordinates of the UE’s positions can be estimated with a back propagation neural network (BPNN) which combines the outputs of DBN and LSTMN. Finally, extensive simulation results are provided to demonstrate the effectiveness of the proposed BPNN scheme. [ABSTRACT FROM AUTHOR]

The time-consuming modeling of physical remote sensing models restricts their application to parameter estimation from satellite observations. Machine learning techniques have become highly developed in recent years and show good capacity for model fitting. Based on our previously developed coupled soil–canopy–atmosphere radiative transfer model (RTM) and a multiple parameters estimation scheme, this paper evaluates the performance of four machine learning algorithms [Gaussian process regression (GPR), back-propagation neural networks (NNs), random forest regression, and general regression NN] on emulating the coupled RTM, where the traditional lookup table (LUT) algorithm is also compared. The results show that the GPR algorithm can emulate complex RTMs with excellent accuracy and efficiency. GPR emulators of photosynthetically active radiation (PAR), fraction of absorbed PAR, and incident shortwave radiation were applied to the multi-parameter estimation scheme to replace the traditional LUT algorithm, which avoids the need to integrate over the spectra while achieving an acceleration ratio of 16. A test of the updated multi-parameter estimation scheme at the Bondville site using 18 years of clear-sky observations demonstrates that replacing the computationally expensive integration processes with GPR emulators is practical. The emulators can also be used to simulate the corresponding parameters independently, and this GPR acceleration method for complex models is universal and can be easily applied to other time-consuming models. [ABSTRACT FROM AUTHOR]

User churn is an important issue in online services that threatens the health and profitability of services. Most of the previous works on churn prediction convert the problem into a binary classification task where the users are labeled as churned and non-churned. More recently, some works have tried to convert the user churn prediction problem into the prediction of user return time. In this approach which is more realistic in real world online services, at each time-step the model predicts the user return time instead of predicting a churn label. However, the previous works in this category suffer from lack of generality and require high computational complexity. In this paper, we introduce ChOracle, an oracle that predicts the user churn by modeling the user return times to service by utilizing a combination of Temporal Point Processes and Recurrent Neural Networks. Moreover, we incorporate latent variables into the proposed recurrent neural network to model the latent user loyalty to the system. We also develop an efficient approximate variational inference algorithm for learning parameters of the proposed RNN by using back propagation through time. Finally, we demonstrate the superior performance of ChOracle on a wide variety of real world datasets. [ABSTRACT FROM AUTHOR]

This paper focuses on improving indoor Wi-Fi localization by mitigating the effect of fluctuation of received signal strength indication (RSSI). The RSSI data collected at each reference point is first transformed through translation and scaling. The BP (Back Propagation) neural network is then used to construct the ranging model using the transformed RSSI to determine the distances between the target point and each reference point. A genetic algorithm (GA) is developed to optimize the initial values of weights and biases of the BP neural network. For convenience, our proposed ranging model is denoted as GTBPD. A new localization algorithm is then proposed, which uses the GTBPD model and the sequential quadratic programming (SQP, an iterative nonlinear optimization algorithm), and the algorithm is denoted as GTBPD-LSQP for simplicity. Experiments were conducted in three areas of two different teaching buildings with complex environments. The performance of the proposed GTBPD-LSQP algorithm is evaluated and compared with four existing algorithms. The experimental results show that our proposed GTBPD-LSQP algorithm achieves significantly higher location accuracy than the four existing algorithms. [ABSTRACT FROM AUTHOR]

This paper reports the estimation of the electric field from a large aperture cavity that generates a higher order mode by backpropagation of fields measured in free space. A higher order mode of a TE6,2 mode at 94 GHz is designed and fabricated using a quasi-optical manner for the analysis. The backpropagating field at the aperture is compared with the measured field. The field estimated at the aperture is used to analyze the mode information, such as the amount of mode mixture of two different rotating fields and the phase difference between two modes. This paper provides a quantitative way to analyze the rotating cylindrical modes generated from an oversized cavity. [ABSTRACT FROM PUBLISHER]

In general, an image attribute is a human-nameable visual property that has a semantic connotation. Appropriate modeling of the intrinsic contextual correlations among attributes plays a fundamental role in attribute detection. In this paper, we consider image attribute detection from the perspective of regularized deep learning. In particular, we propose a regularized deep belief network (rDBN) to perform the image attribute detection task, which is composed of two parts: 1) a detection DBN (dDBN) that models the joint distribution of images and their corresponding attributes, which acts as an attribute detector and 2) a contextual restricted Boltzmann machine that explicitly models the correlations among attributes acting as a regularizer that restraints the output detection result given by the dDBN to meet the contextual prior of attributes. Furthermore, we propose an efficient fine-tuning scheme that can further optimize the performance of the dDBN by backpropagation. Experimental results show that the proposed rDBN obtains improvements over the state-of-the-art methods for attribute detection on the benchmark data sets. [ABSTRACT FROM PUBLISHER]

As an electroluminescentdevice, the coupling relationship between light-emitting diode (LED) input currents, optical power, and LED junction temperature is a complicated multiphysics process. In this paper, a simplified LED photoelectron-thermal (PET) model by artificial neural network (ANN), which can translatemultiphysics field issue into a singlephysics field problem, ismentioned to study the coupling relationship. In the first, an LED lumens, optical power, and electric power at different temperatures are monitored in a temperature controlling integrating sphere. Then, a backpropagation (BP) ANN is trained by these data to construct an LED Photo-Electron-Thermal (PET) relationship. In addition, LED luminaire thermal analyzing is performed using a finite-element method on the outputs of the BP ANN. Finally, the advantage of this method in terms of saving computing resources and computing time is analyzed by comparing the degrees of freedom in different models. The result shows that at least 6.7 times computing resource are saved by this method, which will reduce the LED thermal management system analyzing time greatly. Finally, the application and extension of this model are discussed. [ABSTRACT FROM PUBLISHER]

In this paper, we propose a deep convolutional neural network (DCNN) architecture that addresses the problem of video/image concept annotation by exploiting concept relations at two different levels. At the first level, we build on ideas from multi-task learning, and propose an approach to learn concept-specific representations that are sparse, linear combinations of representations of latent concepts. By enforcing the sharing of the latent concept representations, we exploit the implicit relations between the target concepts. At a second level, we build on ideas from structured output learning and propose the introduction, at training time, of a new cost term that explicitly models the correlations between the concepts. By doing so, we explicitly model the structure in the output space (i.e., the concept labels). Both of the above are implemented using standard convolutional layers and are incorporated in a single DCNN architecture that can then be trained end-to-end with standard back-propagation. Experiments on four large-scale video and image data sets show that the proposed DCNN improves concept annotation accuracy and outperforms the related state-of-the-art methods. [ABSTRACT FROM AUTHOR]

In this paper, we propose the novel principal backpropagation networks (PBNets) to revisit the backpropagation algorithms commonly used in training two-stream networks for video action recognition. We content that existing approaches always take all the frames/snippets for the backpropagation not optimal for video recognition since the desired actions only occur in a short period within a video. To remedy these drawbacks, we design a watch-and-choose mechanism. In particular, the watching stage exploits a dense snippet-wise temporal pooling strategy to discover the global characteristic for each input video, while the choosing phase only backpropagates a small number of representative snippets that are selected with two novel strategies, i.e., Max-rule and KL-rule. We prove that with the proposed selection strategies, performing the backpropagation on the selected subset is capable of decreasing the loss of the whole snippets as well. The proposed PBNets are evaluated on two standard video action recognition benchmarks UCF101 and HMDB51, where it surpasses the state of the arts consistently, but requiring less memory and computation to achieve high performance. [ABSTRACT FROM AUTHOR]

Security is now becoming a well-established challenge for integrated circuits (ICs). Various types of IC attacks have been reported, including reverse engineering IPs, dumping on-chip data, and controlling/modifying IC operation. IEEE 1149.1, commonly known as Joint Test Action Group (JTAG), is a standard for providing test access to an IC. JTAG is primarily used for IC manufacturing test, but also for in-field debugging and failure analysis since it gives access to internal subsystems of the IC. Because the JTAG needs to be left intact and operational after fabrication, it inevitably provides a “backdoor” that can be exploited outside its intended use. This paper proposes machine learning-based approaches to detect illegitimate use of the JTAG. Specifically, JTAG operation is characterized using various features that are then classified as either legitimate or attack. Experiments using the OpenSPARC T2 platform demonstrate that the proposed approaches can classify legitimate JTAG operation and known attacks with significantly high accuracy. Experiments also demonstrate that unknown and disguised attacks can be detected with high accuracy as well (99% and 94%, respectively). [ABSTRACT FROM AUTHOR]

An in-depth study on driving habits and personalized driving-assistance systems is conducive to the realization of vehicle safety and intelligent driving. In this paper, we present a personalized vehicle lane-change assistance system integrated with a driver-behavior identification strategy. First, the driver-behavior data-acquisition system is designed and established. Based on this, the input data of different kinds of drivers along with vehicle signals are collected under typical working conditions. The drivers are classified utilizing factor analysis and a fuzzy c-means clustering algorithm, and the identification of driver behavior is realized using a backpropagation neural network optimized by a particle swarm optimization algorithm. Then, personalized warning, planning, and control systems are designed for lane changing. The proposed personalized lane-change assistance system can provide more personalized recommendations to the drivers, increasing the potential for more widespread acceptance and use of advanced driver-assistance system. Finally, the correctness of the proposed personalized lane-change system is evaluated by conducting computer simulations and a driver-in-the-loop- simulation under various conditions. And the results show that the lane-change assistance system based on driver behavior can meet the driving needs of different drivers without sacrificing safety. [ABSTRACT FROM AUTHOR]

In this paper, we study a secure cooperative transmission system in which a transmitter (Alice) intends to send a secret message to an authorized receiver (Bob), with the aid of a friendly private cooperative jammer (Oscar), in the presence of a malicious eavesdropper (Eve). Specifically, artificial noise (AN) assisted secrecy beamforming is utilized for secure transmission. Under the imperfect channel state information (CSI) of legitimate channels and the statistic CSI of illegitimate channels, the optimal power allocation ratio between the information-bearing signal and the AN signal for maximizing secrecy rate is derived. Through mathematical analysis, we notice that the high channel estimation error is greatly detrimental to the realization of the positive secrecy rate in secure system. For overcoming this drawback, back propagation neural network (BPNN) is employed for a high accuracy channel estimation to simulate a realistic security communication scenario. The numerical results show that the channel estimation errors result in a decrease in secrecy rate, but the BPNN channel estimator still can guarantee secure transmission that verifies the feasibility of the application of physical layer security. Analytical derivations and numerical simulations are presented to validate the correctness of obtained expressions and our analysis. [ABSTRACT FROM AUTHOR]

A new adaptive backpropagation (BP) algorithm based on Lyapunov stability theory for neural networks is developed in this paper. It is shown that the candidate of a Lyapunov function V(k) of the tracking error between the output of a neural network and the desired reference signal is chosen first, and the weights of the neural network are then updated, from the output layer to the input layer, in the sense that ΔV(κ) = V(κ) — V(κ — 1) < 0. The output tracking error can then asymptotically converge to zero according to Lyapunov stability theory. Unlike gradient-based BP training algorithms, the new Lyapunov adaptive BP algorithm in this paper is not used for searching the global minimum point along the cost-function surface in the weight space, but it is aimed at constructing an energy surface with a single global minimum point through the adaptive adjustment of the weights as the time goes to infinity. Although a neural network may have bounded input disturbances, the effects of the disturbances can be eliminated, and asymptotic error convergence can be obtained. The new Lyapunov adaptive BP algorithm is then applied to the design of an adaptive filter in the simulation example to show the fast error convergence and strong robustness with respect to large bounded input disturbances. [ABSTRACT FROM AUTHOR]

This paper proposes a combined modeling method for switched reluctance machine (SRM) with few samples of flux linkage characteristics measured without rotor clamping devices and position sensors. The proposed method mainly consists of two steps, namely data reconstruction and characteristic description. In data reconstruction, the entire flux linkage characteristics are obtained by training support vector machine (SVM) with the measured few samples. The characteristics from the trained SVM are consistent well with those from the rotor-clamping measurement. In characteristic description, back-propagation neural network (BPNN) is adopted to describe the reconstructed flux linkage characteristics and calculated static torque characteristics. On this basis, the simulation model of the SRM prototype is built in MATLAB. The results from simulation under both motoring and generating mode are compared with those from experiments, and good agreements can be found, which prove the effectiveness of the proposed modeling method. To further demonstrate the accuracy and application of the entire flux linkage characteristics obtained from data reconstruction, BPNN is used again to build the mapping from flux linkage and phase current to rotor position for rotor position estimation, and some results are presented. The applicability of the proposed method to different SRM topologies is discussed as well. [ABSTRACT FROM PUBLISHER]

This paper presents a study of different fuzzy neural network (FNN) learning control methods for brushless dc (BLDC) motor drives. The FNN combines fuzzy logic (FL) with the learning capabilities of an artificial neural network. The study designs an FNN structure and defines four different training algorithms for the FNN, namely, backpropagation (BP), extended Kalman filter (EKF), genetic (GEN), and particle swarm optimization (PSO). These algorithms are examined in the simple application of pattern matching an input set to an output set and determine the strengths and weaknesses of each algorithm. Tests of each learning algorithm by a pattern matching benchmark are achieved via dSPACE DSP MATLAB/Simulink environment and allows for the capability for adaptive self-tuning of the weights and memberships of the input parameters. Thus, this adds a self-learning capability to the initial fuzzy design for operational adaptively and implements the solution on real hardware using a BLDC motor drive system. The success of the adaptive FNN-controlled BLDC motor drive system is verified by experimental results. Testing results show that the EKF method is the superior method of the four for this specific application. The BP method was also somewhat successful, nearly matching the pattern but not to the accuracy of the EKF. The GEN and PSO methods did not demonstrate success. Demonstrating the proposed self-learning FNN control on real hardware realizes the solution. [ABSTRACT FROM AUTHOR]

This paper presents a new approach for compensating nonlinear errors of a load cell, based on the optimal neural network with an augmented Lagrange multiplier (ALMNN). The load cell has serious nonlinear errors, which lower the accuracy of the weighing results. In this proposed approach, first, we construct the constraints for training the neural network (NN) using the load cell’s prior knowledge, i.e., the monotonic increasing property of a load cell’s input-output function. Second, we create the augmented Lagrange function with a multiplier and a penalty factor to serve as an NN’s performance index, and then deduce the ALMNN’s detailed training algorithm. The penalty factor plays an important role in ALMNN, thus the reasonable range of numerical value is discussed in this paper. Meanwhile, the weighing principle of a strain gauge load cell is introduced and its nonlinear error model is established. Finally, the experimental results demonstrate the effectiveness of ALMNN. By comparing with the conventional data induction method (DINN, i.e., the training of NN not with the prior knowledge but only the data samples), ALMNN significantly improves the NN’s generalization ability, especially in the case of lacking training samples. In addition, the applicability of ALMNN is verified by three additional experiments. [ABSTRACT FROM PUBLISHER]

Prosperity of the online keyword auctions greatly facilitates the penetration of search-engine marketing in various industries. However, the current operation rules of the search engine make it very difficult for those inexperienced advertisers to make sound bids without the support of powerful tools. Therefore, many studies have been conducted to help advertisers understand such dynamic, infinite, and opaque auction situation, and obtain as good as possible bidding result. This paper, focusing on predicting the return on investment (ROI) of a keyword portfolio, develops a hierarchical Bayesian network (BN) model to forecast keyword auctions’ performance. Few papers directly predict the ROI of a keyword portfolio. This approach effectively echoes advertisers’ expectation for a keyword auction by choosing the right keywords and bids to achieve the desired outcome. The building blocks of the prediction model, such as bid and rank, are organized in a tree-shaped structure with a set of joint conditional probabilities. To infer the posterior probabilities of the predictors, a Bayesian parameter-learning algorithm is conducted after validating the network's structural relationships. The empirical study demonstrates that the prediction model is appropriate and effective for keyword auctions. Moreover, the proposed hierarchical BN model shows a higher accuracy than the popular prediction approach—the back-propagation artificial neural network. [ABSTRACT FROM PUBLISHER]

In this paper, we propose a novel detector for single-channel long-haul coherent optical communications, termed stochastic digital backpropagation (SDBP), which takes into account noise from the optical amplifiers in addition to handling deterministic linear and nonlinear impairments. We discuss the design approach behind this detector, which is based on the maximum a posteriori (MAP) principle. As closed-form expressions of the MAP detector are not tractable for coherent optical transmission, we employ the framework of Bayesian graphical models, which allows a numerical evaluation of the proposed detector. Through simulations, we observe that by accounting for nonlinear signal—noise interactions, we achieve a significant improvement in system reach with SDBP over digital backpropagation (DBP) for systems with periodic inline optical dispersion compensation. In uncompensated links with high symbol rates, the performance difference in terms of system reach for SDBP over DBP is small. In the absence of noise, the proposed detector is equivalent to the well-known DBP detector. [ABSTRACT FROM PUBLISHER]

Remaining useful life (RUL) prediction of lithium-ion batteries can assess the battery reliability to determine the advent of failure and mitigate battery risk. The existing RUL prediction techniques for lithium-ion batteries are inefficient for learning the long-term dependencies among the capacity degradations. This paper investigates deep-learning-enabled battery RUL prediction. The long short-term memory (LSTM) recurrent neural network (RNN) is employed to learn the long-term dependencies among the degraded capacities of lithium-ion batteries. The LSTM RNN is adaptively optimized using the resilient mean square back-propagation method, and a dropout technique is used to address the overfitting problem. The developed LSTM RNN is able to capture the underlying long-term dependencies among the degraded capacities and construct an explicitly capacity-oriented RUL predictor, whose long-term learning performance is contrasted to the support vector machine model, the particle filter model, and the simple RNN model. Monte Carlo simulation is combined to generate a probabilistic RUL prediction. Experimental data from multiple lithium-ion cells at two different temperatures is deployed for model construction, verification, and comparison. The developed method is able to predict the battery's RUL independent of offline training data, and when some offline data is available, the RUL can be predicted earlier than in the traditional methods. [ABSTRACT FROM AUTHOR]

A quasi-continuous high magnetic field is a special kind of pulsed magnetic field with a flat-top field longer than 100 ms. This paper proposes a hybrid power supply that consists of a capacitor power supply and a generator-rectifier power supply. A dual-coil magnet is powered to generate a 65 T/100 ms quasicontinuous high magnetic field with a ripple less than 500 ppm at the Wuhan National High Magnetic Field Center, Wuhan, China. A protection circuit consists of an R-C branch and a diode array is discussed for the worst short-circuit fault, which occurs between the first turns of inner coil and outer coil. This type of faultwill produce oscillation voltage with a high peak value and high frequency on the rectifiers. A self-adaptive PI controller with the coefficients corrected by a back propagation neural network is adopted to reduce the ripple of a flat-top. [ABSTRACT FROM AUTHOR]

Dissolved gas analysis (DGA) of insulating oil can provide an important basis for transformer fault diagnosis. To improve diagnosis accuracy, this paper presents a new transformer fault diagnosis method based on deep belief networks (DBN). By analyzing the relationship between the gases dissolved in transformer oil and fault types, the Non?code ratios of the gases are determined as the characterizing parameter of the DBN model. DBN adopts multi-layer and multi-dimension mapping to extract more detailed differences of fault types. In this process, the diagnosis parameters are pre-trained. A back-propagation algorithm adjusts them with the labels of the samples and optimizes the parameters. To verify the effect of the proposed method, the diagnostic DBN model is constructed and tested using various oil chromatographic datasets collected from the State Grid Corporation of China and previous publications. The performances of the DBN diagnosis model are analyzed by different characterizing parameters, different training datasets and sample datasets. In addition, the influence of discharge and overheating multiple faults on the diagnosis model is studied. The performance of the proposed approach is compared with that derived from support vector machine (SVM), back-propagation neural network (BPNN) and ratio methods respectively. The results show that the proposed method significantly improves the accuracy of power transformer fault diagnosis. [ABSTRACT FROM PUBLISHER]

In this paper, we present and evaluate a new propagation model for heterogeneous networks. The designed model is multiband, multienvironment, and is usable for short and long distance. For this research, a measurement campaign was conducted in Tunis (Tunisia) using continuous wave analog technology. It concerns the most used bands (450, 850, 1800, 2100, and 2600 MHz) in rural, suburban, and urban environments. Measurements are split into two independent and random sets. The first one is used for model training, whereas the second is used for model validation. The new model is based on neural networks, uses back propagation algorithm, and obtains its inputs from Standard Propagation Model, to which we have added more parameters such as frequency, environment type, land use distribution, and diffraction loss. Model variables are computed from accurate Digital Terrain Model and Land Used maps with 2-m resolution. The statistical analysis has shown that the developed model is accurate as we obtained the following metrics: 0.235-dB absolute mean error, 6.850-dB standard deviation, and 85% correlation factor. The obtained simulation results are then compared to SPM and ITU-R P.1812-4 prediction, which are taken as reference to highlight the benefit of the new model. [ABSTRACT FROM AUTHOR]

In this paper, a novel calibration algorithm based on real-valued time-delay neural network (RVTDNN) is proposed for the multiport receiver to recover I/Q signals from the received modulated signals, especially from wideband signals. All the system impairments, including the nonideal six-port correlator, the nonlinearity of the diodes, and the memory effects under the excitation of the wideband signals, are taken into consideration. The RVTDNN, which has three layers, consists of four inputs and two outputs (in-phase and quadrature signals), and its parameters (biases and weights) are all real-valued and are determined by the backpropagation algorithm. Finally, the proposed calibration method is validated by a 2–18-GHz six-port receiver under 3G and 4G signals. The performance of the error vector magnitude between the transmitted signals and received signals is always less than 1.8%, which validates the correctness of the proposed model. [ABSTRACT FROM PUBLISHER]

Neural networks (NNs) have numerous applications to online processes, but the problem of stability is rarely discussed. This is an extremely important issue because, if the stability of a solution is not guaranteed, the equipment that is being used can be damaged, which can also cause serious accidents. It is true that in some research papers this problem has been considered, but this concerns continuous-time NN only. At the same time, there are many systems that are better described in the discrete time domain such as population of animals, the annual expenses in an industry, the interest earned by a bank, or the prediction of the distribution of loads stored every hour in a warehouse. Therefore, it is of paramount importance to consider the stability of the discrete-time NN. This paper makes several important contributions. 1) A theorem is stated and proven which guarantees uniform stability of a general discrete-time system. 2) It is proven that the backpropagation (BP) algorithm with a new time-varying rate is uniformly stable for online identification and the identification error converges to a small zone bounded by the uncertainty. 3) It is proven that the weights' error is bounded by the initial weights' error, i.e., overfitting is eliminated in the proposed algorithm. 4) The BP algorithm is applied to predict the distribution of loads that a transelevator receives from a trailer and places in the deposits in a warehouse every hour, so that the deposits in the warehouse are reserved in advance using the prediction results. 5) The BP algorithm is compared with the recursive least square (RLS) algorithm and with the Takagi–Sugeno type fuzzy inference system in the problem of predicting the distribution of loads in a warehouse, giving that the first and the second are stable and the third is unstable. 6) The BP algorithm is compared with the RLS algorithm and with the Kalman filter algorithm in a synthetic example. [ABSTRACT FROM AUTHOR]

This paper presents and evaluates artificial neural network (ANN) models used for macrocell path-loss prediction. Measurement data obtained by utilizing the IS-95 pilot signal from a commercial code-division multiple-access (CDMA) mobile network in rural Australia are used to train and evaluate the models. A simple neuron model and feed-forward networks with different numbers of hidden layers and neurons are evaluated regarding their training time, prediction accuracy, and generalization properties. Furthermore, different backpropagation training algorithms, such as gradient descent and Levenberg--Marquardt, are evaluated. The artificial neural network inputs are chosen to be distance to base station, parameters easily obtained from terrain path profiles, land usage, and vegetation type and density near the receiving antenna. The path-loss prediction results obtained by using the ANN models are evaluated against different versions of the semi-terrain based propagation model Recommendation ITU-R P.1546 and the Okumura--Hata model. The statistical analysis shows that a non-complex ANN model performs very well compared with traditional propagation models with regard to prediction accuracy, complexity, and prediction time. The average ANN prediction results were 1) maximum error: 22 dB; 2) mean error: 0 dB; and 3) standard deviation: 7 dB. A multilayered feedforward network trained using the standard backpropagation algorithm was compared with a neuron model trained using the Levenberg--Marquardt algorithm. It was found that the training time decreases from 150 000 to 10 iterations, while the prediction accuracy is maintained. [ABSTRACT FROM AUTHOR]

With the development of distributed computing environment, many learning problems now have to deal with distributed input data. To enhance cooperations in learning, it is important to address the privacy concern of each data holder by extending the privacy preservation notion to original learning algorithms. In this paper, we focus on preserving the privacy in an important learning model, multilayer neural networks. We present a privacy-preserving two-party distributed algorithm of backpropagation which allows a neural network to be trained without requiring either party to reveal her data to the other. We provide complete correctness and security analysis of our algorithms. The effectiveness of our algorithms is verified by experiments on various real world data sets. [ABSTRACT FROM AUTHOR]

Bounding ellipsoid (BE) algorithms offer an attractive alternative to traditional training algorithms for neural networks, for example, backpropagation and least squares methods. The benefits include high computational efficiency and fast convergence speed. In this paper, we propose an ellipsoid propagation algorithm to train the weights of recurrent neural networks for nonlinear systems identification. Both hidden layers and output layers can be updated. The stability of the BE algorithm is proven. [ABSTRACT FROM AUTHOR]