Multivariate time series (MTS) anomaly detection is essential to identify faults and unexpected patterns. However, existing methods often struggle with capturing both temporal dependencies and variable correlations while addressing the high dimensionality and heterogeneity of MTS data. To tackle these challenges, we propose V2AD, a dual-view graph representation learning framework for MTS anomaly detection. V2AD decomposes temporal and variable modeling into two complementary views, leveraging patch-based graph modeling to capture multi-scale dependencies efficiently. An adversarial training strategy en-forces consistency within normal data and amplifies discrepancies between normal and anomalous regions. Extensive experiments on four benchmark datasets demonstrate that V2AD outperforms state-of-the-art methods, offering superior anomaly detection accuracy and scalability. This work provides a robust solution to the complexities of MTS data and sets a foundation for real-time and broader time-series applications.

In nature, fish locomotion is primarily classified into the BCF (body and caudal fin) propulsion mode and the MPF (median and paired fin) propulsion mode. This paper presents a bio-inspired robotic electric ray that integrates a BCF-mode caudal fin with MPF-mode pectoral fins. The caudal fin consists of a set of wire-driven, multi-joint active segments coupled with a soft, compliant segment, while each symmetrical pectoral fin incorporates two sets of wire-driven joints and a soft fin structure. The undulatory motion of the MPF-mode pectoral fins enables the robotic ray to execute maneuvers such as forward swimming, backward swimming, and in-place turning, whereas the BCF-mode caudal fin enhances linear swimming and turning capabilities. Experimental results demonstrate that MPF-mode swimming achieves a maximum speed of 0.190 m/s (0.358 BL/s), while the cooperative propulsion of MPF and BCF modes enables speeds of up to 0.352 m/s (0.664 BL/s). Notably, the robotic electric ray's large pectoral fins can function as grippers, allowing it to grasp and transport objects using caudal fin propulsion, thereby facilitating object manipulation tasks.

In the context of the information age, the frequent occurrence of software vulnerabilities has emerged as a critical issue demanding immediate resolution. Traditional vulnerability detection methods have struggled to keep pace with the escalating security demands, while the advent of deep learning technology has introduced novel solutions to software vulnerability detection. Deep learning not only facilitates the automatic extraction of features, thereby reducing the cost of manual intervention, but also demonstrates remarkable advantages across various domains. In recent years, research on vulnerability detection based on deep learning has achieved notable progress, yet it still faces several limitations. This study focuses on C/C++ program vulnerability detection and proposes an enhanced approach, VulSCS, based on secondary slicing. By performing secondary slicing on source code exhibiting vulnerable behaviors, this method extracts code segments with higher representational value, thereby capturing richer vulnerability-related features. Experimental results indicate that, compared to state-of-the-art vulnerability detection tools, VulSCS improves detection accuracy by 3.2% and enhances detection efficiency by approximately threefold. This research offers new perspectives and methodologies for deep learningbased software vulnerability detection.

The large petrochemical pelletizer gearbox is a critical component in synthetic chemicals, accurate forecasting of its time series from online oil monitoring is of utmost importance for safe operation. However, existing time series forecasting methods face challenges in handling oil monitoring scenarios characterized by disturbances, multimodality, and large time spans. To address these issues, we propose an intelligent forecasting method named multiscale inverted transform network (MITN). First, the multimodal online oil monitoring time series collected from moisture sensors, viscosity sensors, abrasive image sensors, and metal particle sensors are utilized to perform correlation analysis and identify the key nonlinear variables. In addition, the multiscale module is further employed to obtain comprehensive redundant characteristics. Second, the inverted perspective is designed for modeling from the time dimension and the variable dimension. Third, the multiple multiattention mechanism module is utilized with the feedforward network to learn the semantic meaning of variable time series and produce accurate forecasting results. Finally, layer normalization is used to improve the training stability and convergence to eliminate the distribution difference between variables. We used online oil monitoring time series from a petrochemical pelletizer gearbox from July 2018 to May 2023 for validation. The results show that MITN not only can obtain smaller forecasting errors than the existing time series forecasting networks, such as long short-term memory, gated recurrent unit, temporal convolutional network, and transform, but also can effectively generalize to the unknown variables. The proposed MITN pioneers ideal multivariate time series forecasting for complex online oil monitoring, with the potential to enhance the operational safety of large petrochemical plants.

The sports industry occupies an important position in China's national economy, and the sports event industry is its core industry, and its role in regional economic development has received more and more attention from all parties. This paper takes Sichuan Province as an example to discuss the role of sports events in regional development. Firstly, it is a theoretical overview. This paper explores the mechanism of its economic effect from four perspectives: economic growth, industrial structure, regional agglomeration and city brand. Second is the case study. Through the grey correlation analysis of the relationship between sports and tourism industry, using the exponential smoothing method and ARMA method, and establishing a multi-indicator model, it is concluded that sports and tourism industry have the highest correlation and play a positive role in promoting regional development.

The existing fault diagnosis of harmonic drives is difficult to deploy vibration sensors, and the diagnosis accuracy is insufficient only using current signal. Therefore, we propose an intelligent fault diagnosis method using a multimodal collaborative meta network (MCMN) with severely missing modality. First, multimodal data of the harmonic drive were collected to analyze. Second, a feature reconstruction network is used to achieve a unified model to handle missing modalities in testing. Besides, the uncertainty assessment is used as a feature regularization network to overcome data bias. Third, the inference of the meta-learning is used to obtain the generated weights, and carried out the joint optimization. Finally, MCMN is optimized by using the approximate lower bound of Monte Carlo. Experimental results show that the accuracy of MCMN is 93.75% in the case of full-modalities training data and the severely missing modalities testing data, which is better than the existing method.

Fish locomotion which adopts body and/or caudal fin swimming mode consists of different motions, such as Cruising-straight, Cruising-turn, and various fast turns, among others. Currently, there is no single mathematical model that could illustrate all these motions. Thus, for scientists and engineers, it is quite cumbersome and complicated to model and control different motions with multiple principles. In this article, we proposed a general kinematic model to illustrate the kinematics of all aforementioned swimming motions. The model is synthesized by a nonlinear oscillator and a traveling wave equation. By changing four parameters extracted from the model, the kinematic model can demonstrate all the aforementioned swimming motions with different amplitudes and frequencies. To verify the model, we built a multijoint robotic fish and developed its dynamic model and control method to perform all the maneuvers under the guidance of the general kinematic model. Through this systematic methodology, one can easily study the principles of different swimming motions and design the multimotions controller for a robotic fish through only one governing kinematic model.

Benefiting from the promising performance of CNNs models for high-level vision tasks, these networks have been extensively adopted to image enhancement tasks. However, recent methods have complex architecture resulting in poor generalization and high computational cost. Their activation functions are originally designed for other vision tasks. In this work, we present a lightweight network to learn periodic features (LPF) using the proposed wave presentation. Specifically, to better capture implicit feature representations, we represent features as signals with three parts: Cosine Wave Map (CWM), Sine Wave Map (SWM) and Direct Current Map (DCM). Thus, we formulate the image enhancement task as a signal modulation problem. Inspired by the Fourier transform, we build the Fourier Enhancement Module (FEM) that allows for efficient and scalable spatial mixing of local and non-local contents and dynamically learns the interaction between waves to enhance the images. LPF with only 80k parameters achieves better quantitative and qualitative results compared with SOTA methods on four image enhancement datasets. The source code and pretrained model are available at https://github.com/DeniJsonC/LPF.

In an effort to further advance semi-supervised generative and classification tasks, we propose a simple yet effective training strategy called *dual pseudo training* (DPT), built upon strong semi-supervised learners and diffusion models. DPT operates in three stages: training a classifier on partially labeled data to predict pseudo-labels; training a conditional generative model using these pseudo-labels to generate pseudo images; and retraining the classifier with a mix of real and pseudo images. Empirically, DPT consistently achieves SOTA performance of semi-supervised generation and classification across various settings. In particular, with one or two labels per class, DPT achieves a Fréchet Inception Distance (FID) score of 3.08 or 2.52 on ImageNet $256\times256$. Besides, DPT outperforms competitive semi-supervised baselines substantially on ImageNet classification tasks, *achieving top-1 accuracies of 59.0 (+2.8), 69.5 (+3.0), and 74.4 (+2.0)* with one, two, or five labels per class, respectively. Notably, our results demonstrate that diffusion can generate realistic images with only a few labels (e.g., $<0.1$%) and generative augmentation remains viable for semi-supervised classification. Our code is available at *https://github.com/ML-GSAI/DPT*.

Video snapshot compressive imaging (SCI) captures multiple sequential video frames by a single measurement using the idea of computational imaging. The underlying principle is to modulate high-speed frames through different masks and these modulated frames are summed to a single measurement captured by a low-speed 2D sensor (dubbed optical encoder); following this, algorithms are employed to reconstruct the desired high-speed frames (dubbed software decoder) if needed. In this article, we consider the reconstruction algorithm in video SCI, i.e., recovering a series of video frames from a compressed measurement. Specifically, we propose a Spatial-Temporal transFormer (STFormer) to exploit the correlation in both spatial and temporal domains. STFormer network is composed of a token generation block, a video reconstruction block, and these two blocks are connected by a series of STFormer blocks. Each STFormer block consists of a spatial self-attention branch, a temporal self-attention branch and the outputs of these two branches are integrated by a fusion network. Extensive results on both simulated and real data demonstrate the state-of-the-art performance of STFormer. The code and models are publicly available at https://github.com/ucaswangls/STFormer.

Abstract We present a simulation microscope device called GiantScope, which combines virtual microscope, cloud computing, and embedded technologies. Users can complete most of the microscope‐based experiments in biology courses by operating our device, while learning the operating skills of microscopes at the same time. Our device supports most of the operation functions of optical microscopes, including quasifocus screw adjustment, slide movement recognition and so on, and also has auxiliary enhancement functions including manual measurement, annotation and so on. In addition, we have built a cloud‐based digital slide database, which enables users to select experimental observations through digital slides, including static cell specimens or dynamic cell activities. After user study, we found that using GiantScope for biological experiments has better learning efficiency and user experience than traditional microscopes.

Abstract In view of the problem that the traditional learning service recommendation does not fully consider the distinct differences between individuals, it is easy to lead to the contradiction between unchanging learning resources and learners’ personalized learning needs that are constantly improving, so an adaptive learning service recommendation improvement algorithm based on big data is proposed. Idea is based on adaptive learning platform and function modules. We consider the individual differences between students, to students as the center, collect students’ personalized learning demand data, and according to the data information to build student demand model. On the basis of using data mining methods for clustering recommendation service resources in learning, the adaptive recommend according to students’ individual need is proposed. The experimental results show that the adaptive learning service recommendation algorithm based on big data has high recommendation accuracy, coverage rate and recall rate, which is of great significance in the actual learning service recommendation.

Cellular network performance depends heavily on the setting of its network parameters. Current practice of parameter setting relies largely on expert experience, which is often suboptimal and time-consuming. Therefore, how to find the optimal parameters automatically is increasingly concerned by network operator. This article proposes a collaborative learning approach to predict the performance of network elements under different parameters based on factorization machine. The proposed approach captures the potential time-aware correlation between network elements and their parameters to boost the accuracy and efficiency of performance prediction. Extensive experiment results demonstrate that the proposed approach can significantly improve the effectiveness of performance prediction, based on a large-scale real-world network dataset collected from a metropolitan LTE network.

Soft robot has demonstrated promise in unstructured and dynamic environments due to unique advantages, such as safe interaction, adaptiveness, easy to actuate, and easy fabrication. However, the highly dissipative nature of elastic materials results in small stiffness of soft robot which limits certain functions, such as force transmission, position accuracy, and load capability. In this paper, we present a novel articulated soft robot with variable stiffness. The robot is constructed by rigid joints and compliant bistable structures in series. Each joint can be independently locked through triggering the bistable structure to touch the mechanical constrain. Thus, the bending stiffness of the joint can be magnified which increases the stiffness of the articulated soft robot. Through this construction method, even driven by only one servomotor, the robot demonstrates variable workspace and stiffness which have the potential of dexterous manipulation and maintaining shape under tip load.

With the feature size shrinking down to 7 nm and beyond, the impact of wire resistance is significantly growing, and the circuit delay incurred by metal wires is noticeably raising. To address this issue, a new technique called via pillar insertion is developed. However, the poor success rate of the via pillar insertion process immediately becomes an important problem. In this paper, we explore the causes of via pillar insertion failures by experiments on the ISPD 2015 benchmarks, which are embedded with a real industrial cell library. The results show that the reasons for the low success rate may be due to track misalignment, power and ground stripe overlapping, and insufficient margin area. Therefore, we propose the first detailed placement flow which is aware of via pillars to maximize the success rate of via pillar insertion. In the proposed flow, we first filter out infeasible cell rows and then move the via pillar-inserting cells to their eligible positions. Next, we adopt a two-stage legalization method with high flexibility on cell ordering based on a dynamic programming-based detailed placement algorithm. Finally, we improve congested rows with a global moving process. Experiment results show that our algorithm improves the insertion rates by 54-58%, and achieves over 99% insertion rate on average.

At the end of digital integrated circuit (IC) design flow, some nets may still be left open due to engineering change order (ECO). Resolving these opens could be quite challenging for some huge nets such as power ground nets because of a large number of obstacles and greatly distributed net components. Existing studies on multilayer obstacle-avoiding rectilinear Steiner trees may not be applicable to solve this problem because they assume the pins of an input net is a set of points, while the discrete net components in this problem can be regarded as a set of rectilinear pins. In this paper, we develop an efficient open-net connector that can deal with rectilinear pins. The proposed algorithm flow minimizes the total connection cost based on precise estimation of the shortest distance between each pair of rectilinear net components with the presence of complex obstacles. The experimental results show that the proposed flow can outperform the top-three teams of 2017 CAD contest at ICCAD and a state-of-the-art work in terms of total connection cost or runtime efficiency.

With the dramatically increase of design complexity and the advance of semiconductor technology nodes, huge difficulties appear during design for manufacturability with existing lithography solutions. Sub-resolution assist feature (SRAF) insertion and optical proximity correction (OPC) are both inevitable resolution enhancement techniques (RET) to maximize process window and ensure feature printability. Conventional model-based SRAF insertion and OPC methods are widely applied in industrial application but suffer from the extremely long runtime due to iterative optimization process. In this paper, we propose the first work developing a deep learning framework to simultaneously perform SRAF insertion and edge-based OPC. In addition, to make the optimized masks more reliable and convincing for industrial application, we employ a commercial lithography simulation tool to consider the quality of wafer image with various lithographic metrics. The effectiveness and efficiency of the proposed framework are demonstrated in experimental results, which also show the success of machine learning-based lithography optimization techniques for the current complex and large-scale circuit layouts.

At the end of digital integrated circuit (IC) design flow, some nets may still be left open due to engineering change order (ECO). Resolving these opens could be quite challenging for some huge nets such as power ground nets because of a large number of obstacles and greatly distributed net components. Existing studies on multilayer obstacle-avoiding rectilinear Steiner trees may not be applicable to solve this problem because they assume the pins of an input net is a set of points, while the discrete net components in this problem can be regarded as a set of rectilinear pins. In this paper, we develop an efficient open-net connector that can deal with rectilinear pins. The proposed algorithm flow minimizes the total connection cost based on precise estimation of the shortest distance between each pair of rectilinear net components with the presence of complex obstacles. Experimental results show that the proposed flow can outperform the top three teams of 2017 CAD Contest at ICCAD in terms of total connection cost or runtime efficiency.

Summary Current rapid increasing implementations in data diversity, autonomy, and dynamic privilege management, fine‐grained access controls in social networks have resulted in various challenges in applying existing access control models. The intercrossing relations lead to the complex access control system, which often brings risks when the system is updated or expanded. The implementations of cloud computing has further complicate the access controls due to multiple tenancies and service providers. We focus on this issue and propose a new social network access control model using logical authorization language, named as RuleSN, which can be efficiently used in cloud systems. This model provides high performance of authorization expressiveness and flexibility that can effectively describe relations of User to User (U2U), User to Resource (U2R), Resource to Resource (R2R) and attributes of users and resources. First, this paper elaborates the formal definitions of the RuleSN model. Second, we describe the model's authorization specification and verification policies and explain the syntax and semantics of the authorization language. Finally, the implementation, application, and expressiveness of the model discussed by examples. Copyright © 2016 John Wiley & Sons, Ltd.

The convenience of sharing of digital contents in social network makes protection of copyrighted digital content a concerned problem. So many strict methods such as DRM technology are used to protect digital content, which makes enforcement of properly copyright laws difficult in social network such as fair use , a general concept denoting the legally protected right of people to use content based on exceptions and limitations of copyright laws. The paper firstly discusses the problem of fair use in social network from point of view of laws, rights holders and users of digital content. Then a fair use mechanism based on MRuleSN model, a multi-party authorization model for social networks proposed by the author. Finally, an example demonstrates the effects and flexibility of the methods.

This paper presents a novel robot fish with vector propulsion. It can swim like a shark and/or a dolphin. The propulsor (tail) of the robot has an underactuated serpentine backbone and the actuation is done by two sets of orthogonally distributed wires. The backbone is composed of seven vertebras and an elastic rod. The vertebras are articulated by the rod and spherical joints. The horizontal flapping and vertical flapping are independently actuated by two motors. This enables the propulsor providing thrust in all directions. Propulsion model of the propulsor is developed by integrating the kinematic model and Lighthill's elongated body theory. A prototype is built. Tests show that the robot fish could flap its tail like the shark or the dolphin effectively. In the swimming tests, the maximum swimming speed of the robot is 0.35 BL/s.