Sparse Property Research Articles

Modified locally joint sparse marginal embedding and wasserstein generation adversarial network for bearing fault diagnosis

Rolling bearings are essential parts for manufacturing machines. Vast quantities of features are often extracted from measured signals to comprehensively reflect the conditions of bearings, which may cause high dimensionality, information redundancy, and time consumption. In addition, it is extremely difficult, expensive, and time-consuming to collect samples with label information during the bearing fault diagnosis in real-world scenarios. In this study, a novel bearing defect diagnosis method for small sample size is proposed based on modified local joint sparse marginal embedding (MLJSME) and Wasserstein generative adversarial networks (WGANs). MLJSME can effectively extract intrinsic sparse discriminant features of high-dimensional dataset by preserving both global and local structures. Graph embedding and Gaussian kernel function are adopted to preserve the locality structure of dataset. The global structure and discriminate information are preserved by maximum margin criterion which can also avoid small sample-size problem. Moreover, joint sparsity is applied to preserve the sparse property and improve the robustness to noise and outliers. An abundance of artificial samples can be obtained with WGAN and a few labeled samples. Firstly, a high-dimensional feature dataset consisting of time-domain and frequency-domain features is extracted from original vibration signals, then MLJSME is utilized to extract sensitive low-dimensional features, and a small number of low-dimensional features are fed into WGAN to generate a large number of artificial samples that used to train the classifier, and the bearing fault types can be finally identified. The effectiveness and feasibility of the proposed method is validated by analyzing the different experimental cases.

Unsupervised feature selection regression model with nonnegative sparsity constraints

Selecting appropriate features can better describe the characteristics and structure of data, which play an important role in further improving models and algorithms whether for supervised or unsupervised learning. In this paper, a new unsupervised feature selection regression model with nonnegative sparse constraints (URNS) is proposed. The algorithm combines nonnegative orthogonal constraint, L2,1-norm minimum optimization and spectral clustering. Firstly, the linear regression model between the features and the pseudo labels is given, and the indicator matrix, which describes feature weight, is subject to nonnegative and orthogonal constraints to select better features. Secondly, in order to reduce redundant and even noisy features, L2,1-norm for indicator matrix is added to the regression model for exploring the correlation between pseudo labels and features by the row sparsity property of L2,1-norm. Finally, pseudo labels of all samples are established by spectral clustering. In order to solve the regression model efficiently and simply, the method of nonnegative matrix decomposition is used and the complexity of the given algorithm is analysed. Moreover, a large number of experiments and analyses have been carried out on several public datasets to verify the superiority of the given model.

On sparsity, power-law, and clustering properties of graphex processes

AbstractThis paper investigates properties of the class of graphs based on exchangeable point processes. We provide asymptotic expressions for the number of edges, number of nodes, and degree distributions, identifying four regimes: (i) a dense regime, (ii) a sparse, almost dense regime, (iii) a sparse regime with power-law behaviour, and (iv) an almost extremely sparse regime. We show that, under mild assumptions, both the global and local clustering coefficients converge to constants which may or may not be the same. We also derive a central limit theorem for subgraph counts and for the number of nodes. Finally, we propose a class of models within this framework where one can separately control the latent structure and the global sparsity/power-law properties of the graph.

Optimizing Tensor Programs on Flexible Storage

Tensor programs often need to process large tensors (vectors, matrices, or higher order tensors) that require a specialized storage format for their memory layout. Several such layouts have been proposed in the literature, such as the Coordinate Format, the Compressed Sparse Row format, and many others, that were especially designed to optimally store tensors with specific sparsity properties. However, existing tensor processing systems require specialized extensions in order to take advantage of every new storage format. In this paper we describe a system that allows users to define flexible storage formats in a declarative tensor query language, similar to the language used by the tensor program. The programmer only needs to write storage mappings, which describe, in a declarative way, how the tensors are laid out in main memory. Then, we describe a cost-based optimizer that optimizes the tensor program for the specific memory layout. We demonstrate empirically significant performance improvements compared to state-of-the-art tensor processing systems.

RIS-Aided mmWave MIMO Channel Estimation Using Deep Learning and Compressive Sensing

Reconfigurable intelligent surface (RIS) assisted wireless systems require accurate channel state information (CSI) to control wireless channels and improve both the bandwidth and energy efficiency. However, CSI acquisition is non-trivial for two reasons: 1) the passive nature of RIS does not allow transceiving and processing pilot signals, and 2) the dimensions of the cascaded channel between transceivers increases with the large number of RIS elements, which yields high training overhead and computational complexity. While prior art has mainly focused on frequency-flat channel estimation, this paper proposes novel data-driven and compressive sensing based approaches for estimating both frequency-flat and frequency-selective cascaded channels of RIS-assisted multi-user millimeter-wave large multiple input multiple output (MIMO) systems with limited training overhead. The proposed methods exploit the common sparsity property among the different subcarriers and the double-structured sparsity property of the angular cascaded channel matrices as different angular cascaded channels observed by different users share completely common non-zero rows and user-specific column supports. The proposed data-driven cascaded channel estimation approaches use denoising neural networks to accurately detect channel supports. Alternatively, when data-training capabilities are not available, the compressive sensing based orthogonal matching pursuit (OMP) approach relies on sparsity properties and applies simultaneous OMP to detect the channel supports. Simulation results show that the pilot overhead required by the proposed scheme is lower than existing schemes. When compared to other OMP approaches that achieve an NMSE gap of 5 to 6 dB with respect to the Oracle least square lower bound, the proposed algorithms reduce the lower bound gap to only 1 dB, while reducing complexity by more than two orders of magnitude.

Wind Power Prediction using Multi-Task Gaussian Process Regression with Lagged Inputs

Wind is a renewable energy source that has become more important in recent years. Wind turbines are equipped with a SCADA system, which allows for remote supervision of the wind farm. SCADA systems are customarily used to provide data averaged every 10 minutes. Nevertheless, recent literature suggests that more insights could be extracted with a higher granularity of data. In this work, a naive methodology based on Multi-Task Gaussian Process Regression is presented, in order to show how spatiotemporal modeling benefits power estimation. Using sparsity properties a model for possible power prediction is proposed. The model proposed performs better than the power curves provided by the manufacturer.

Accelerating distributed machine learning with model compression and graph partition

The rapid growth of data and parameter sizes of machine learning models makes it necessary to improve the efficiency of distributed training. It is observed that the communication cost usually is the bottleneck of distributed training systems. In this paper, we focus on the parameter server framework which is a widely deployed distributed learning framework. The frequent parameter pull, push, and synchronization among multiple machines leads to a huge communication volume. We aim to reduce the communication cost for the parameter server framework. Compressing the training model and optimizing the data and parameter allocation are two existing approaches to reducing communication costs. We jointly consider these two approaches and propose to optimize the data and parameter allocation after compression. Different from previous allocation schemes, the data sparsity property may no longer hold after compression. It brings additional opportunities and challenges for the allocation problem. We also consider the allocation problem for both linear and deep neural network (DNN) models. Fixed and dynamic partition algorithms are proposed accordingly. Experiments on real-world datasets show that our joint compression and partition scheme can efficiently reduce communication overhead for linear and DNN models.

Substructural damage identification using autoregressive moving average with exogenous inputs model and sparse regularization

Substructuring approaches possess many superiorities over traditional global approaches in damage identification because large-size global structures are replaced by small and manageable substructures. This paper proposes a substructural time series model for locating and quantifying the damage in complex systems. A substructural autoregressive moving average with exogenous inputs (ARMAX) model is established to extract the frequencies and mode shapes of substructures as indicators for damage detection. The detection of structural damage is essentially an inverse problem, and the damage in structure bears sparse properties. The inverse problem of substructural damage identification is efficiently solved via sparse regularization, and structural damage can be located and quantified through the nonzero terms in the solution vector. The accuracy of the proposed method is demonstrated by the numerical simulation of a frame structure and shaking table test of a shear building structure. As the substructural properties are more sensitive to local structural damage than the global properties, the substructural ARMAX model is quite accurate and efficient to be used in the damage identification of a complex system.

A robust mixed error coding method based on nonconvex sparse representation

Linear representation based methods have been extensively applied in image recognition, especially for those with noise, illumination changes, and occlusions. However, most existing methods assume a specific distribution for image noise estimation, which is intractable to handle complex variations. Besides, they usually use convex norm to describe the noise sparse and low-rank property, and it is a biased approximation. To address these problems, we propose a novel nonconvex regularized robust mixed error coding (NRRM) method, which uses mixed norms from both 1D and 2D perspectives to model the complex image noise without convex relaxation. In specific, we use weighted ℓ2-norm based robust coding to characterize the sparse noise in images, and weighted matrix nuclear norm to characterize the low-rank noise. Compared with traditional regression approaches, our method can more fine-grained and accurate to capture noise and alleviate its negative influence for robust recognition. Besides, we constrain the representation component in a group-wise manner to weigh the roles of different classes. The NRRM model is solved efficiently by adopting an alternating direction method of multipliers (ADMM) algorithm. Comprehensive experiments on some benchmark face image databases validate the superiority of NRRM over several state-of-the-art linear representation based methods.

Isolation Forest Based Submodule Open-Circuit Fault Localization Method for Modular Multilevel Converters

Fault localization is one of the most important issues for modular multilevel converters (MMCs) consisting of numerous switches. This article proposes an isolation forest (IF) based submodule (SM) switch open-circuit fault localization method for MMCs. Based on the continuous sampling SM capacitor voltages, a number of isolation trees (ITs) are produced to construct the IFs for MMCs. Through the comparison of continuous IFs’ outputs, the faulty SM can be effectively localized. The proposed IF-based fault localization method only requires SM capacitor voltages in the MMC to construct concise low-data-volume tree models, and uses sparsity and difference properties of outlier data to localize fault, and accordingly it simplifies calculation complexity. In addition, it does not require the MMC's mathematical models and manual setting of empirical thresholds. Simulation and experiment are conducted, and the results confirm the effectiveness of proposed method.

An adaptive group sparse feature decomposition method in frequency domain for rolling bearing fault diagnosis

Group-sparse mode decomposition (GSMD) is a decomposition method designed based on the group sparse property of signals in frequency domain. It is proved to be highly efficient and robust against noise, showing bright prospects for bearing fault diagnosis. However, the following adverse factors may impede its application for incipient bearing fault feature extraction: Initially, the GSMD method did not consider the impulsiveness and periodicity of the bearing fault feature. As a result, the ideal filter bank generated by GSMD may not accurately cover the fault frequency band because it may produce over-coarse or over-narrow filter bank under the condition of strong interference harmonics, large random shocks and heavy noise. Moreover, the location of informative frequency band was obstructed since the bearing fault signal shows complicated distribution in frequency domain. To overcome the abovementioned limitations, an adaptive group sparse feature decomposition (AGSFD) method is proposed. Firstly, the harmonics, large-amplitude random shocks and periodic transient feature are modeled as limited bandwidth signals in the frequency domain. On this basis, an autocorrection of envelope derivation operator harmonic to noise ratio (AEDOHNR) indicator is proposed to guild the construction and optimization of the filter bank of AGSFD. Also, the regularization parameters of AGSFD are adaptively determined. With the optimized filter bank, the original bearing fault is decomposed into a serial of components with AGSFD method, where the sensitive fault-induced periodic transient component is maintained using the AEDOHNR indicator. Finally, the studies of the simulation and two experimental items are carried out to evaluate the feasibility and the superiority of AGSFD method. The results indicate the AGSFD method can identify the early failure in the presence of heavy noise, strong harmonics or random shocks and has better decomposition efficiency.

Full-field single molecule localization microscopy with a mono-detector.

Fluorescence super-resolution microscopy has reached the ultimate detection limit of individual molecules. The sparsity property has enabled a change of paradigm from a diffraction-limited resolution in the few hundred nanometer range to a molecular localization-based resolution in the few nanometer range. Usually, single molecule localization microscopies are based on PSF fitting using a camera detector. Alternatively, another localization strategy is possible based on the dependence of the fluorophore emission on the illumination. A sequence of acquisition using different structured illuminations can be used to localize the emitter. This excitation-based strategy has been applied successfully for wide-field images using a shifting modulated illumination synchronized with a camera (ModLoc, SIMFlux, ROSE,…) and in a scanning approach configuration (MinFlux) with a single detector. However, these localization techniques are limited by the slowness of the camera in their full-field configuration and by the scanning approach in their mono-detector configuration. In this work, we present a novel concept for molecular localization microscopy where the information of the entire full-field image is given by the emitted photon flux. Although working as a full-field technique, this allows emitter localization via the acquisition of the fluorescence light by a single photon counting module, rather than a camera, which has the immediate advantage of gaining temporal resolution. The theoretical resolution limit (given by its Cramér-Rao lower bounds) gives a better localization precision as compared to the standard PSF fitting. We will show experimental characterization of this new localization technique in terms of spatial precision and temporal performances. Imaging of calibrated samples, as well as cytoskeleton network in COS7 cells will be presented, and the different assets of this new approach will be fully discussed.

Achieving Reliable Intervehicle Positioning Based on Redheffer Weighted Least Squares Model Under Multi-GNSS Outages.

Achieving reliable intervehicle positioning is one of the most fundamental elements for many vehicular applications, including collision avoidance and autonomous driving. Vehicle position is generally provided by a global navigation satellite system (GNSS), which unfortunately suffers from inaccuracy to varying degrees in challenging environments, for example, GNSS outages. In this article, a reliable fusion technique, called non-Gaussian Redheffer weighted least squares ( n GRWLSs), is proposed. This new approach highlights the intervehicle positioning estimation in multi-GNSS outage environments, such as complete, partial, and free GNSS pseudorange outages. The proposed method combines, on the one hand, the benefits of the Gaussian dynamical matrix principle and the Redheffer distribution function for the sparse property in complete GNSS pseudorange outages and, on the other hand, the use of the optimal window size to regulate the data flow generated by both the inertial navigation systems (INSs) and GNSS during a partial GNSS pseudorange outage. During the free GNSS pseudorange outage, the process ignores data from the INS, and instead, GNSS pseudorange information alone will be considered to compute the intervehicle positioning information. Consequently, weighted least squares is used as an intervehicle positioning estimator. To address the pseudorange uncommon and INS measurement noises, the generalized error distribution (GED) is used to estimate the non-Gaussian densities. Finally, road-test experiments are implemented to evaluate the consistency of the proposed approach. The experimental results show that the proposed n GRWLS can accurately estimate the intervehicle positioning under various conditions (free, partial, and complete GNSS pseudorange outages).

Projection domain processing for low-dose CT reconstruction based on subspace identification.

Low-dose computed tomography (LDCT) has promising potential for dose reduction in medical applications, while suffering from low image quality caused by noise. Therefore, it is in urgent need for developing new algorithms to obtain high-quality images for LDCT. This study tries to exploit the sparse and low-rank properties of images and proposes a new algorithm based on subspace identification. The collection of transmission data is sparsely represented by singular value decomposition and the eigen-images are then denoised by block-matching frames. Then, the projection is regularized by the correlation information under the frame of prior image compressed sensing (PICCS). With the application of a typical analytical algorithm on the processed projection, the target images are obtained. Both numerical simulations and real data verifications are carried out to test the proposed algorithm. The numerical simulations data is obtained based on real clinical scanning three-dimensional data and the real data is obtained by scanning experimental head phantom. In simulation experiment, using new algorithm boots the means of PSNR and SSIM by 1 dB and 0.05, respectively, compared with BM3D under the Gaussian noise with variance 0.04. Meanwhile, on the real data, the proposed algorithm exhibits superiority over compared algorithms in terms of noise suppression, detail preservation and computational overhead. The means of PSNR and SSIM are improved by 1.84 dB and 0.1, respectively, compared with BM3D under the Gaussian noise with variance 0.04. This study demonstrates the feasibility and advantages of a new algorithm based on subspace identification for LDCT. It exploits the similarity among three-dimensional data to improve the image quality in a concise way and shows a promising potential on future clinical diagnosis.

Improving Generalizability of Spectral Reflectance Reconstruction Using L1-Norm Penalization

Spectral reflectance reconstruction for multispectral images (such as Weiner estimation) may perform sub-optimally when the object being measured has a texture that is not in the training set. The accuracy of the reconstruction is significantly lower without training samples. We propose an improved reflectance reconstruction method based on L1-norm penalization to solve this issue. Using L1-norm, our method can provide the transformation matrix with the favorable sparse property, which can help to achieve better results when measuring the unseen samples. We verify the proposed method by reconstructing spectral reflection for four types of materials (cotton, paper, polyester, and nylon) captured by a multispectral imaging system. Each of the materials has its texture and there are 204 samples in each of the materials/textures in the experiments. The experimental results show that when the texture is not included in the training dataset, L1-norm can achieve better results compared with existing methods using colorimetric measure (i.e., color difference) and shows consistent accuracy across four kinds of materials.

Ultra-Massive MIMO Channel Measurements at 5.3 GHz and a General 6G Channel Model

Ultra-massive multiple-input multiple-output (MIMO) technology will bring unique channel characteristics that need to be fully explored through channel measurements and channel modeling. In this paper, single-user and multi-user channel measurements using ultra-massive MIMO antenna arrays with different configurations are conducted at 5.3 GHz band. The non-stationarity, spherical wavefront, channel hardening, and sparse properties are validated by the channel measurements. Correspondingly, a general three-dimensional (3D) sixth generation (6G) non-stationary geometry-based stochastic model (GBSM) for ultra-massive MIMO communication systems is proposed. The statistical properties of channel measurements and the corresponding channel model are studied, including delay power spectral density (PSD), angular PSD, spatial cross-correlation function (SCCF), normalized user-side correlation matrix, singular value spread (SVS), degrees of freedom (DoF), and diversity level. In addition, channel capacities of channel measurements and the corresponding channel model are studied. The accuracy of the proposed general channel model is validated by the consistency of simulation results and measurement results, which indicates that the proposed model can be applied to ultra-massive MIMO communication systems.

Transmission Defects Localization Network: Towards Wrong Assembly in the Transmission Assembly Process

Transmission valve component is an important part of automobile manufacture. The success of the assembly of transmission valve is directly related to driving safety of vehicles. While localizing transmission assembly defects is particularly important in assembly of transmission valve component. As an image processing problem, real-time assembly images of transmission valve component are adopted to determine whether the assembly is correct or wrong. Transmission valve component in these images have small, severe reflection, and sparse properties, which increases the difficulty of detection. Therefore, this paper proposes a transmission defects localization network based on Siamese network for improving the performance of assembly of transmission valve components. In our model, we establish an image similarity evaluation network with designed multi-scale features fusion approach. Furthermore, in order to reduce intra-class spacing by similar transmission valve part samples on evaluation action, an improved binary cross entropy and focal loss function is discovered for feature re-processing. Finally, experimental results on real-world transmission assembly dataset indicate that our proposed approach outperforms other compared methods.

Deep Atomic Norm Denoising Network for Jointly Range-Doppler Estimation With FRI Sampling

Finite rate of innovation (FRI) sampling is widely used in modern high-resolution range-Doppler (RD) detection systems to reduce the sampling rate. However, as the sampling rate decreases, the method becomes more noise-sensitive. This study proposes a deep atomic norm denoising network (DAND-Net) to denoise FRI samples of RD signals and estimate the noise levels (NLs). Based on the sparse common support property of the FRI samples, the denoising problem of the RD signals was modeled as an atomic norm soft thresholding problem, which jointly used multiple measurement vectors and could be unfolded into a model-driven deep network. The trainability and generalization ability of the network were improved by introducing piecewise linear functions and new trainable variables. By introducing an NL estimation operator, the NL could be better estimated during denoising, and the network training and convergence better supported. Additionally, the plug-and-play method could be used to complete denoising prior to many existing FRI methods. The simulation experiments showed that the proposed denoising network can achieve better denoising and NL estimation results with fewer iterations.

Sparse Low-Rank Matrix Estimation With Nonconvex Enhancement for Fault Diagnosis of Rolling Bearings

The diagnosis of bearing early fault is significant and fundamental in the machine condition monitoring. The accurate and effective diagnosis is of great importance to avoid further serious accidents. However, existing sparse low-rank (SLR) methods for bearing fault diagnosis suffer from underestimation of amplitude and inaccurate approximation of singular values. Therefore, in this paper, a novel sparse low-rank matrix estimation method with nonconvex enhancement (SLRNE) is proposed, extracting the fault transients from observed noisy signal. Specifically, fault transients have both sparse and low-rank properties in time-frequency domain. Based on this, a SLR optimization model is proposed to simultaneously promote the above two properties via truncated nuclear norm (TNN) and generalized minimax concave (GMC) penalty function. The two nonconvex functions aim to promote low-rank property and sparsity respectively. Then, based on derived convexity conditions of the optimization problem, convex optimization algorithm, alternating direction method of multipliers (ADMM) and forward-and-backward splitting (FBS) algorithm, are applied to obtain global optimal solution. In the iterative algorithm, a weighting strategy is designed for the singular value threshold operator to enhance the effect of fault feature extraction. Simulated and experimental signals verify the effectiveness of SLRNE and contrast experiments verify its superiority.

Target Imaging Using Compressed Sampling in Synthetic Aperture Interferometric Radiometer

The target imaging application is significant to various sensing systems, such as radiometers, radars, and infrared. However, high system complexity impedes the application of interferometric radiometers to target imaging tasks to some extent. Specifically for an <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">N</i> -element interferometric radiometer with aperture synthesis technique, complex correlators are of the order of O( <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">N</i> <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> ), giving rise to the great difficulty of system hardware implementation. In this paper, we propose a new compressed interferometric radiometer (CIR) concept for target imaging applications, which exploits the sparsity property of targets in the spatial domain. The CIR target imaging framework mainly adopts the compressive measurement method to acquire partial visibility function samples in the spatial-frequency domain via a proper sparse sampling pattern. Then, these partially observed visibility samples are inverted to image the target contrast information by sparse recovery methods. For the above image recovery process, we propose two novel algorithms named local regional information-based reweighted <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</i> 1-norm minimization (LRRL1) and local regional convolution-based reweighted <italic xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">l</i> 1-norm minimization (LRCRL1). The experiments using simulated and real data demonstrate the validity and effectiveness of the proposed CIR target imaging framework, showing superiority in both imaging performance and system complexity compared with conventional algorithms used in interferometric radiometers.