Unsupervised domain adaptation (UDA) is becoming a prominent solution for the domain-shift problem in many time-series classification tasks. With sequence properties, time-series data contain both local and sequential features, and the domain shift exists in both features. However, conventional UDA methods usually cannot distinguish those two features but mix them into one variable for direct alignment, which harms the performance. To address this problem, we propose a novel virtual-label-based hierarchical domain adaptation (VLH-DA) approach for time-series classification. Specifically, we first slice the original time-series data and introduce virtual labels to represent the type of each slice (called local patterns). With the help of virtual labels, we decompose the end-to-end (i.e., signal to time-series label) time-series task into two parts, i.e., signal sequence to local pattern sequence and local pattern sequence to time-series label. By decomposing the complex time-series UDA task into two simpler subtasks, the local features and sequential features can be aligned separately, making it easier to mitigate distribution discrepancies. Experiments on four public time-series datasets demonstrate that our VLH-DA outperforms all state-of-the-art (SOTA) methods.

Multi-Intent Spoken Language Understanding (SLU), a novel and more complex scenario of SLU, is attracting increasing attention. Unlike traditional SLU, each intent in this scenario has its specific scope. Semantic information outside the scope even hinders the prediction, which tremendously increases the difficulty of intent detection. More seriously, guiding slot filling with these inaccurate intent labels suffers error propagation problems, resulting in unsatisfied overall performance. To solve these challenges, in this paper, we propose a novel Scope-Sensitive Result Attention Network (SSRAN) based on Transformer, which contains a Scope Recognizer (SR) and a Result Attention Network (RAN). SR assignments scope information to each token, reducing the distraction of out-of-scope tokens. RAN effectively utilizes the bidirectional interaction between SF and ID results, mitigating the error propagation problem. Experiments on two public datasets indicate that our model significantly improves SLU performance (5.4% and 2.1% on Overall accuracy) over the state-of-the-art baseline.

Spoken Language Understanding (SLU), a core component of the task-oriented dialogue system, expects a shorter inference facing the impatience of human users. Existing work increases inference speed by designing non-autoregressive models for single-turn SLU tasks but fails to apply to multi-turn SLU in confronting the dialogue history. The intuitive idea is to concatenate all historical utterances and utilize the non-autoregressive models directly. However, this approach seriously misses the salient historical information and suffers from the uncoordinated-slot problems. To overcome those shortcomings, we propose a novel model for multi-turn SLU named Salient History Attention with Layer-Refined Transformer (SHA-LRT), which comprises a SHA module, a Layer-Refined Mechanism (LRM), and a Slot Label Generation (SLG) task. SHA captures salient historical information for the current dialogue from both historical utterances and results via a well-designed history-attention mechanism. LRM predicts preliminary SLU results from Transformer’s middle states and utilizes them to guide the final prediction, and SLG obtains the sequential dependency information for the non-autoregressive encoder. Experiments on public datasets indicate that our model significantly improves multi-turn SLU performance (17.5% on Overall) with accelerating (nearly 15 times) the inference process over the state-of-the-art baseline as well as effective on the single-turn SLU tasks.

Spoken Language Understanding (SLU), a core component of the task-oriented dialogue system, expects a shorter inference latency due to the impatience of humans. Non-autoregressive SLU models clearly increase the inference speed but suffer uncoordinated-slot problems caused by the lack of sequential dependency information among each slot chunk. To gap this shortcoming, in this paper, we propose a novel non-autoregressive SLU model named Layered-Refine Transformer, which contains a Slot Label Generation (SLG) task and a Layered Refine Mechanism (LRM). SLG is defined as generating the next slot label with the token sequence and generated slot labels. With SLG, the non-autoregressive model can efficiently obtain dependency information during training and spend no extra time in inference. LRM predicts the preliminary SLU results from Transformer's middle states and utilizes them to guide the final prediction. Experiments on two public datasets indicate that our model significantly improves SLU performance (1.5% on Overall accuracy) while substantially speed up (more than 10 times) the inference process over the state-of-the-art baseline.

Spoken language understanding (SLU), which is a core component of the task-oriented dialogue system, has made substantial progress in the research of single-turn dialogue. However, the performance in multi-turn dialogue is still not satisfactory in the sense that the existing multi-turn SLU methods have low portability and compatibility for other single-turn SLU models. Further, existing multi-turn SLU methods do not exploit the historical predicted results when predicting the current utterance, which wastes helpful information. To gap those shortcomings, in this paper, we propose a novel Result-based Portable Framework for SLU (RPFSLU). RPFSLU allows most existing single-turn SLU models to obtain the contextual information from multi-turn dialogues and takes full advantage of predicted results in the dialogue history during the current prediction. Experimental results on the public dataset KVRET have shown that all SLU models in baselines acquire enhancement by RPFSLU on multi-turn SLU tasks.

In recommender systems, users' preference and items' attraction are heavily determined by users' and items' attributes information. The aim of this study was to incorporate these attributes information into a matrix factorization model. In this paper, users' and items' attributes were used to calculate similarity, and then a nonparametric Bayesian method was applied to cluster users and cluster items, based on the similarity. These clustering results were used to mine coupling relationships and improve the accuracy of recommendation. Experimental results on Movielens 1M demonstrate the superiority of our proposed model over existing methods.

Reconstructing the sparse signal from a few linear measurements has attracted increasing attentions in recent years. In this study, the authors propose the alternating projection (AP) method for sparse signal recovery with learning the sparsity of the original signal. Different with classical hard thresholding algorithms, the AP method regards the signal recovery problem as finding an intersect point of two sets. Theoretically, the authors prove that the proposed algorithm can reconstruct the s ‐sparse original signal provided the sensing matrix satisfies several assumptions when the noise is absent. They also prove that AP method is a noise‐robust algorithm, i.e. a tolerable reconstruction can be obtained by AP if the noise is small. In numerical experiments, the authors compare AP with several existing algorithms when being applied to sparse signals recovery and images reconstruction. The results demonstrate the efficiency of the proposed algorithm.

The nonsmooth and nonconvex regularization has many applications in imaging science and machine learning research due to its excellent recovery performance. A proximal iteratively reweighted nuclear norm algorithm has been proposed for the nonsmooth and nonconvex matrix minimizations. In this paper, we aim to investigate the convergence of the algorithm. With the Kurdyka-Łojasiewicz property, we prove the algorithm globally converges to a critical point of the objective function. The numerical results presented in this paper coincide with our theoretical findings.

This article describes a Matlab implementation of a mixed extended-precision package with three multiple-component formats. In the package, double-double, triple-double, and quad-double numbers, which are unevaluated sums of two, three and four IEEE double precision numbers respectively, are defined as Matlab classes including real and complex formats. We implement the four basic operations with mixed extended-precision based on these classes, and apply it to other operations and elementary functions. All operations and functions we presented are overloaded in Matlab. We experimentally evaluate this package that it allows the user to obtains more accurate numerical results with Matlab.

Polynomials are widely used in scientific computing and engineering. In this paper, we present an accurate and fast compensated algorithm to evaluate bivariate polynomials with floating-point coefficients. This algorithm is applying error free transformations to the bivariate Horner scheme and sum the final decomposition accurately. We also prove the proposed algorithm's accuracy with forward error analysis that the accuracy of the computed result is similar to the result computed by the bivariate Horner scheme in twice the working precision. Numerical experiments illustrate the behavior and it has higher efficiency than the bivariate Horner scheme implemented in double-double library.

Designing fast singular value decomposition (SVD) is significantly interesting in applications. The random direct SVD (RSVD) has provided a fast scheme to compute the well-approximate SVD by unilateral randomized sampling. In this paper, we present an efficient random algorithm in a bilateral sampling way. We also prove that the proposed algorithms can be bounded well and have less computational complexity compared to RSVD when the objective matrix is approximately square. Numerical experiments on graph Laplacian and Hilbert matrix demonstrate the efficiency and stability of the proposed methods.

For solving the l 1 ‐ l 1 minimisation problem, the authors propose a reweighted fast iterative shrinkage thresholding algorithm. The proposed algorithm consists of two steps: in the first step, the authors apply the smoothing technique to l 1 ‐ l 1 minimisation; and in the second step the smoothed problem is solved by fast iterative shrinkage thresholding algorithm (FISTA). With the help of restarts technique, the authors further accelerate the reweighted FISTA algorithm. Compared with some provable and efficient existing methods, the methods proposed in this study enjoy faster speed, less parameters and that the convergent analysis does not need any assumption of A . On the computational level, numerical experiments on sparse signal recovery demonstrate the efficiency of the proposed methods.

Fuzzy c‐means (FCM) clustering algorithm has been widely used in image segmentation. In this study, a modified FCM algorithm is presented by utilising local contextual information and structure information. The authors first establish a novel similarity measure model based on image patches and local statistics, and then define the neighbourhood‐weighted distance to replace the Euclidean distance in the objective function of FCM. Validation studies are performed on synthetic and real‐world images with different noises, as well as magnetic resonance brain images. Experimental results show that the proposed method is very robust to noise and other image artefacts.

The authors address the problem of support recovery with multiple measurement vectors (MMV) in this study. The scale of an MMV is reduced by using the singular value decomposition technique, and a novel l 2 – l 1 minimisation model with two variables for the reduced MMV is proposed. Then a new alternating minimisation algorithm based on the alternating direction method of multipliers is presented. They prove the globally convergence property of the presented algorithm. Several numerical simulations both on random data and for direction‐of‐arrival estimation are conducted to evaluate the performance of the proposed method for support recovery of MMV.

This letter addresses the joint sparse recovery problem, which is a hot topic in the compressed sensing (CS) theory and its various applications. Inspired by particle swarm optimization (PSO) algorithm and some sparse recovery algorithms, a novel swarm intelligence algorithm called M-SISR is proposed to solve the problem. In M-SISR, the initial positions of the swarm are designed using theq-thresholding (1 ≤q≤ 2) algorithm, and the update strategy is designed using the ideas of PSO and some sparse recovery algorithms. Theoretical analysis shows the good property of the update strategy, and numerical simulations on random Gaussian data illustrate the efficiency of M-SISR.

This letter addresses the stable signal recovery problem in the presence of noisy perturbations, and presents a variant of Subspace Pursuit (SP) method, called Randomly Enhanced Adaptive SP (REASP) method. REASP first adds correlative and irrelative atoms with an adaptive technique and a random strategy respectively, next detects the overall reliability, and then deletes unreliable atoms per iteration. Through these modifications, REASP is expected to achieve superior performance. The complexity analysis shows that the computational complexity of REASP is a bit higher but acceptable. The experimental results illustrate the superior performance of REASP, compared to the frequently used greedy pursuit (GP) andl1basis pursuit (BP) methods.

The l1-constrained minimal singular value (l1-CMSV) of the sensing matrix for sparse signal recovery is investigated in this letter. By exploiting certain geometrical properties of the (l_1-CMSV), we derive new sufficient conditions for the recovery of both exactly and approximately sparse signals. Moreover, we demonstrate that a class of structured random matrices, including the Fourier random matrices and the Hadamard random matrices, can satisfy these sufficient conditions by showing their l_1-CMSV are bounded away from zero with high probability, as long as the number of measurements is reasonably large.

Multiplicative noise removal based on total variation (TV) regularization has been widely researched in image science. In this paper, inspired by the spatially adapted methods for denoising Gaussian noise, we develop a variational model, which combines the TV regularizer with local constraints. It is also related to a TV model with spatially adapted regularization parameters. The automated selection of the regularization parameters is based on the local statistical characteristics of some random variable. The corresponding subproblem can be efficiently solved by the augmented Lagrangian method. Numerical examples demonstrate that the proposed algorithm is able to preserve small image details, whereas the noise in the homogeneous regions is sufficiently removed. As a consequence, our method yields better denoised results than those of the current state-of-the-art methods with respect to the signal-to-noise-ratio values.

The goal of image quality assessment research is to design quantitative measurements for the evaluation of image quality such that it is consistent with subjective human evaluation. Inspired by intrinsic geometric structure of nature images and characteristic of visual perception, we propose a novel geometric structural distortion model for image quality assessment in this paper, which has relatively low computational complexity and clear physical meanings. The experimental results of LIVE image database show that the proposed method is consistent with the subjective assessment of human beings and has a good performance for all distortion types.

The interscale Stein's unbiased risk estimator (SURE)-based approach introduced by Luisier is a recent state of the art in orthonormal wavelet denoising, but it is not very effective for those images that have substantial high-frequency contents. To solve this problem, we introduce an effective integration of the intrascale correlations within the interscale SURE-based approach. We show that the consideration of both the intrascale and interscale dependencies of wavelet coefficients brings more denoising gains than those obtained with the interscale SURE-based approach, especially for denoising of images that have substantial textures such as the Barbara image.

Based on a lifting scheme and the construction theorem of biorthogonal wavelets, we propose a new symmetric biorthogonal 9/7-tap wavelet called LS97. Compared to Cohen-Daubechies-Feauveau 9/7-tap (CDF 9/7-tap) wavelet adopted by JPEG2000, when the new wavelet is applied to image coding, the compression performance is exactly the same as that of the CDF 9/7-tap wavelet, while computational complexity is reduced remarkably.

A type of lapped biorthogonal transform that can map integers to integers is considered. Based on the computational structure of the proposed lapped transform, an efficient multiplierless algorithm for the lapped biorthogonal transform (LBT), called integer lapped biorthogonal transform (IntLBT), is proposed. The proposed IntLBT is implemented by a series of dyadic lifting steps, and provides fast, efficient computation of the transform coefficients as well as the ability to map integers to integers. Application of the novel IntLBT in lossy image coding gains very competitive results comparing to the performance of the much more complex Cohen-Daubechies-Feauveau (CDF) 9/7-tap biorthogonal wavelet with irrational coefficients.