Deep transfer learning (DTL), which incorporates new ideas from deep neural networks into (TL), has achieved excellent success in computer vision, text classification, behavior recognition, and natural language processing. As a branch of machine learning, DTL applies end-to-end to overcome the drawback traditional that regards each dataset individually. Although some valuable impressive general surveys exist on TL, special attention recent advances are lacking. In this survey, we first...

With the rapid development of science and technology, high-dimensional data have been widely used in various fields. Due to complex characteristics data, it is usually distributed union several low-dimensional subspaces. In past decades, subspace clustering (SC) methods studied as they can restore underlying perform fast with help self-expressiveness property. The SC aim construct an affinity matrix by self-representation coefficient then obtain results using spectral method. key how design...

Unsupervised feature selection (UFS) is widely applied in machine learning and pattern recognition. However, most of the existing methods only consider a single sparsity, which makes it difficult to select valuable discriminative subsets from original high-dimensional set. In this paper, we propose new UFS method called DSCOFS via embedding double sparsity constrained optimization into classical principal component analysis (PCA) framework. Double refers using $\ell_{2,0}$-norm $\ell_0$-norm...

Multi-view clustering leverages consistent and complementary information across multiple views to provide more comprehensive insights than analysis of single-view data. However, the heterogeneity redundancy high-dimensional mixed multi-view data pose significant challenges existing techniques. In this paper, we propose a novel fusion regularized method with adaptive group sparsity, enabling reliable while effectively capturing local features. Technically, for features exhibiting different...

In this paper, we propose a fast and convergent algorithm to solve unassigned distance geometry problems (uDGP). Technically, construct novel quadratic measurement model by leveraging $\ell_0$-norm instead of $\ell_1$-norm in the literature. To nonconvex model, establish its optimality conditions develop iterative hard thresholding (IHT) algorithm. Theoretically, rigorously prove that whole generated sequence converges L-stationary point with help Kurdyka-Lojasiewicz (KL) property. Numerical...

Anomaly detection (AD) is an important topic in remote sensing, aiming to identify unusual or abnormal features within the data. However, most existing low-rank representation methods usually use nuclear norm for background estimation, and do not consider different contributions of singular values. Besides, they overlook spatial relationships regions, particularly failing fully leverage 3D structured information Moreover, noise practical scenarios can disrupt structure background, making it...

In order to improve the performance of canonical correlation analysis (CCA) based methods for process monitoring, this brief proposes a novel monitoring approach using structured joint sparse (SJSCCA). Technically, graph Laplacian could incorporate variable information and sparsity discard useless variables. The developed two-stage alternating direction method multipliers is shown be very efficient because each derived subproblem has closed-form solution or can solved by fast solvers. detect...

This article proposes a novel modeling method for the stochastic nonlinear degradation process by using relevance vector machine (RVM), which can describe nonlinearity of more flexibly and accurately. Compared with existing methods, where processes are modeled as Wiener drift function formulized power law or exponential law, this kind characterize behavior. Instead coefficient directly, weighted combination basis functions is utilized to express increment parameters calculated sparse...

In order to achieve more accurate predicted RUL in the early stage of degradation, a novel remaining useful life (RUL) prediction method for stochastic degradation process is proposed. Technically, modeling as Wiener (WP) whose drift increment weighted sum kernel functions can flexibly depict nonlinear trend. Introducing long short term memory (LSTM) network capture long-term dependencies offline experimental and online observed data forecast future increment. Then, based on model, numerical...

As modern industrial processes become complicated, and some faults are difficult to be detected due noises nonlinearity of data, data-driven fault detection (FD) has been extensively used detect abnormal events in functional units. To obtain better FD performance nonnegative matrix factorization (NMF), this article first proposes an method using the structured joint sparse orthogonal NMF (SJSONMF). The core idea is incorporate graph regularization, sparsity, orthogonality constraints into...

Canonical correlation analysis (CCA) has attracted great interest in multi-view representation. However, most of the CCA methods heavily rely on matrix structure, which may neglect prior geometric information high-order data. To deal with above issue, we first propose a novel tensor formulation orthogonality, called TCCA-O, based Tucker decomposition to preserve orthogonality. Then, incorporate structured sparse regularization term into TCCA-OS, improve feature In addition, develop an...

Nonnegative matrix factorization (NMF) is an efficient dimension reduction technique, which has been extensively used in the fields, such as image processing, automatic control, and machine learning. The application to fault detection (FD) still not investigated sufficiently. To improve performance of NMF-based FD approaches, this article proposes a novel approach using structured joint sparse NMF (SJSNMF) for non-Gaussian processes. basic idea SJSNMF incorporate graph Laplacian preserve...

This article proposes an efficient nonlinear process monitoring method (DCCA-SCO) by integrating canonical correlation analysis (CCA), deep autoencoder neural networks (DAENNs), and sparsity-constrained optimization (SCO). Specifically, DAENNs are first used to learn a function automatically, which characterizes intrinsic features of the original data. Then, CCA is performed in that low-dimensional representation space extract most correlated variables. In addition, SCO imposed reduce...

Multifocus image fusion has attracted considerable attention because it can overcome the physical limitations of optical imaging equipment and fuse multiple images with different depths field into one full-clear image. However, most existing deep learning-based methods concentrate on segmentation focus–defocus regions, resulting in loss details near boundaries. To address issue, this article proposes a novel generation adversarial network dense connections (Fusion-UDCGAN) to...

The fusion of light detection and ranging (LiDAR) inertial measurement unit (IMU) sensing information can effectively improve the environment modeling localization accuracy navigation systems. To realize spatiotemporal unification data collected by IMU LiDAR, a two-step calibration method combining coarse fine is proposed. mainly includes two aspects: (1) Modeling continuous-time trajectories attitude motion using B-spline basis functions; LiDAR estimated normal distributions transform (NDT)...

Abstract With the development of industrial intelligence, data-driven fault diagnosis plays an important role in prognostics and health management. However, there is usually a large amount unlabeled data from different working conditions, making cross-domain unstable inflexible. To deal with this issue, we propose two novel transfer subspace learning methods based on low-rank sparse representation (LRSR), called LRSR-G LRSR-R. Specifically, integrates additional matrix LRSR to characterize...

Sparse linear discriminant analysis (LDA) is a popular machine learning method that improves the accuracy of data classification by introducing sparsity. However, its performance often degrades seriously when encountering noise. To address this issue, paper proposes new called efficient and robust sparse (ERSLDA). The core idea to characterize local pixel corruptions integrating <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML"...

Principal component analysis (PCA) has shown its high efficiency in process monitoring (PM). However, most of the existing PCA-based PM approaches only consider spatial prior and ignore temporal prior. Therefore, this brief, we propose a novel framework using spatiotemporal PCA (STPCA), which incorporates both priors. Technically, is integrated to preserve cause-effect relationship variables, embedded maintain geometric structure samples. Moreover, an efficient optimization algorithm...

In this article, we propose an efficient fault diagnosis framework to achieve accurate isolation. The core is introduce the <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"><tex-math notation="LaTeX">$\ell _{2,0}$</tex-math></inline-formula> -norm sparsity constrained optimization reduce variable redundancy and determine number, which different from existing sparse variants. order illustrate idea, article takes principal component...

Process monitoring (PM) is important for improving product quality and ensuring plant safety in industrial systems. Recently, canonical correlation analysis (CCA)-based PM has shown excellent performance. The core idea to first seek the relationship between two sets of process variables via CCA, then construct a residual generator determine test statistics, finally achieve best monitoring. However, global local structure not fully exploited. In this article, we propose new nonlinear approach...