One-shot pruning efficiently compresses Large Language Models but produces coarse sparse weights, causing significant performance degradation. Traditional fine-tuning approaches to refine these weights are prohibitively expensive for large models. This highlights the need for a training-free weight refinement method that works seamlessly with one-shot pruning and can efficiently recover the lost performance. To tackle this problem, we propose Efficient Iterative Weight Refinement (EIWR), a lightweight, plug-and-play, and training-free method that refines pruned weights through layer-wise iterative optimization. EIWR achieves efficient weight refinement via three key components: a Global Soft Constraint that eliminates costly row-wise Hessian inversions and expands the solution space; a Historical Momentum Strategy that leverages one-shot pruning priors to accelerate convergence and enhance final performance; and Neumann Series Extrapolation that significantly speeds up per-iteration computation. As a result, EIWR enables effective weight refinement with minimal time and memory overhead. Extensive experiments on LLaMA2/3 and Qwen under different pruning strategies and sparsity levels demonstrate that our method can efficiently refine sparse weights and mitigate performance degradation. For example, on LLaMA2-7B under 70 percent sparsity, EIWR reduces perplexity by 15 percent compared with SparseGPT on the WikiText2 benchmark, with only 1.81 additional minutes of computation and 1GB of additional memory.

The evolution of Internet of Vehicles (IoV) technologies, encompassing wireless communications and Artificial Intelligence (AI), has advanced the collaborative “Pedestrian-Vehicle-Road-Cloud” IoV edge services, enhancing road efficiency and driving safety. Operating in an open-edge environment with vast sensory data, IoV faces significant privacy risks from unauthorized access and data breaches. Consequently, privacy-preserving computation (PPC) is crucial for secure IoV services. This paper reviews PPC techniques in IoV edge services, exploring network characteristics and potential privacy attacks. It categorizes and evaluates techniques such as differential privacy, homomorphic encryption, and secure multi-party computation based on data security, utility, and overhead. Summarizing their pros and cons, the challenges and future research directions for IoV edge services are outlined.

Trusted Execution Environment(TEE) delineates se- cure zones within the system, ensuring the security of the execution environment and data confidentiality. However, due to the relatively late start of the RISC-V instruction set ar- chitecture, there are currently few designs for RISC-V-based TEEs, which face challenges such as performance bottlenecks, insufficient support for privilege modes, and the lack of a unified model. Moreover, synchronous circuits’ reliance on extensive clock networks increases area and power usage. This paper introduces an asynchronous Physical Memory Protection(PMP) design for RISC-V, addressing these issues. This scheme provides key hardware support for implementing TEE for RISC-V. The architecture supports user, supervisor, and machine modes, offering 4 B to 16 GB granular protection and up to 64 PMP entries. The initial address and the number of PMP entries can be configured according to requirements, offering high flexibility. After testing, the overall and sub-module functions are correct and align with expectations. Finally, in the 110-nm process, the area of this architecture is 0.47 mm2and the power consumption is only 3.6829 mW.

Intelligent parking recommendations are essential for enhancing parking space utilization and alleviating traffic congestion. Current parking recommendation systems are mostly based on collecting a large amount of raw parking-related data. However, this often imposes a significant communication and computing burden on resource-constrained edge vehicles and roadside facilities, resulting in high communication overhead, latency, and particularly privacy risks. In this article, a lightweight privacy-aware edge intelligent parking recommendation scheme (FuzzyTop) is proposed. Raw parking data is converted into fuzzy information on the edge sides using intuitionistic fuzzy sets (IFSs), and then a fuzzy multiattribute ranking algorithm is devised to determine the top$k$optimal parking lots for drivers. We evaluated our scheme using real-world data sets collected from Hangzhou and Shanghai in China. The experimental results show that the FuzzyTop algorithms outperform the state-of-the-art benchmarks in terms of data transmission, accuracy, and real-time performance.

As an integral component of the Advanced Driver Assistance System (ADAS), intelligent car following plays a vital role in decreasing the accident rate on hazardous roads. Existing car-following methods have issues such as undesired real-time performance and privacy protection. To this end, this paper proposes a privacy-aware fuzzy prediction of the front car braking and fuzzy decision-making of the rear car braking based on dynamic uncertain traffic conditions. Sensitive data, such as location and speed, are transformed into fuzzy information before transmission. This information is then utilized to predict the braking behavior of the front vehicle. Fuzzy rules are developed to facilitate real-time car-following decisions in a lightweight manner. Extensive experimental results show that the overall prediction accuracy of fuzzyFollow reaches 90.4%, outperforming the state-of-the-art work. The proposed scheme outperforms the compared counterparts in terms of communication cost, real-time performance, and privacy protection.

Look-Up Table (LUT) has recently gained increasing at-tention for restoring High-Quality (HQ) images from Low-Quality (LQ) observations, thanks to its high computational efficiency achieved through a “space for time” strategy of caching learned LQ-HQ pairs. However, incorporating multiple LUTs for improved performance comes at the cost of a rapidly growing storage size, which is ultimately re-stricted by the allocatable on-device cache size. In this work, we propose a novel LUT compression framework to achieve a better trade-off between storage size and performance for LUT-based image restoration models. Based on the observation that most cached LQ image patches are dis-tributed along the diagonal of a LUT, we devise a Diagonal-First Compression (DFC) framework, where diagonal LQ-HQ pairs are preserved and carefully re-indexed to main-tain the representation capacity, while non-diagonal pairs are aggressively subsampled to save storage. Extensive ex-periments on representative image restoration tasks demon-strate that our DFC framework significantly reduces the storage size of LUT-based models (including our new de-sign) while maintaining their performance. For instance, DFC saves up to 90% of storage at a negligible performance drop for x 4 super-resolution. The source code is available on GitHub: https://github.com/leenas233IDFC.

We propose FaceCom, a method for 3D facial shape completion, which delivers high-fidelity results for incomplete facial inputs of arbitrary forms. Unlike end-to-end shape completion methods based on point clouds or voxels, our approach relies on a mesh-based generative network that is easy to optimize, enabling it to handle shape completion for irregular facial scans. We first train a shape generator on a mixed 3D facial dataset containing 2405 identities. Based on the incomplete facial input, we fit complete faces using an optimization approach under image inpainting guidance. The completion results are refined through a post-processing step. FaceCom demonstrates the ability to effectively and naturally complete facial scan data with varying missing regions and degrees of missing areas. Our method can be used in medical prosthetic fabrication and the registration of deficient scanning data. Our experimental results demonstrate that FaceCom achieves exceptional performance in fitting and shape completion tasks. The code is available at https://github.com/dragonylee/FaceCom.git.

General sparse matrix-matrix multiplication (SpGEMM) is a fundamental computational method with wide-ranging applications in scientific simulations, machine learning, and image processing. However, when tackling large-scale SpGEMM, single-core processors fall short in managing the computation-intensive tasks, while multi-core and many-core processors encounter challenges such as complex scheduling, high communication overhead, and substantial energy consumption. Existing platforms for SpGEMM that base on synchronous circuit design require complex state machines to implement parsing mechanisms, and they suffer from power consumption issues associated with clock trees. Therefore, the exploration of asynchronous sparse matrix accelerators is essential to address these issues. This paper presents an asynchronous sparse matrix multiplication accelerator architecture designed for 128×128 sparse matrices, called HSAMM, which is based on the Hybrid-Strassen algorithm and employs the BCSR and BCSC formats for the storage of input and output matrices. HSAMM is validated through prototype implementation and performance evaluation on the FACE-VUP platform. Simulation results demonstrate significant performance enhancements in SpGEMM (sparsity ≤ 0.8%) for int32 data type, achieving average accelerations of 3.2×, 8.2× and 83.3× over the Intel MKL, the Eigen library and the matlab, respectively.

Vehicular trajectory data can be widely used in applications such as traffic prediction and congestion control. However vehicular trajectory data is voluminous and requires significant storage and processing resources, which contradicts the resources-constraint vehicular networks. Existing compression methods suffer either low compression effects or privacy leakage. A privacy-aware Trajectory Compression scheme based on Fuzzy markovian Prediction (TCFP) is proposed in this paper, which consists of two steps of fuzzy compression. The first-step compression is achieved by converting the raw trajectory data into fuzzy information on the edge vehicle sides. Further compression is performed at edge RSUs through fuzzy multi-order Markovian prediction combined with new-devised fuzzy deviation filtering rules. Extensive experimental evaluation based on real-world data sets demonstrates the proposed TCFP scheme achieves desired QoS performance in terms of compression rate, compression time and information loss.

Vehicular Ad Hoc Networks (VANETs) hold the opportunity to help improve safe driving and road efficiency. There are great uncertainty and unpredictability in VANETs due to rapid vehicle movement and changeable road conditions, which makes designing a Quality of Service (QoS) aware data forwarding protocol for VANETs remain a significant challenge. In this paper, A novel fuzzy decision based data forwarding scheme is proposed to meet the QoS requirements of high reliability and low latency. Unlike the existing routing methods, fuzzy information instead of raw vehicular data is used for data forwarding decisions. Besides, fuzzy decision strategies based on fuzzy non-dominate optimization and fuzzy rules are devised for selecting roads at junctions and determining relay vehicles at edge vehicles. Experimental results show that our fuzzy forwarding scheme achieves desirable QoS performance in terms of packet arrival rate, end-to-end delay and communication overhead.

In the field of remote sensing image processing, clouds heavily affect the quality of the remote sensing images and their application potential. Thus, in recent years, with the prevalence of deep learning techniques used in the field of image processing, many methods have been proposed for cloud removal using single remote sensing images. The existing single-image cloud removal methods suffer from poor generalization capabilities that prevent them from being applied to diverse remote sensing images. Thus, a novel method using a multimodal architecture is proposed which provides multiple most likely outputs for the image and selects the best one through perception-based image quality evaluator (PIQE). In addition, adversarial consistency loss is used to replace cycle consistency loss, which encourages the model to retain more texture information of the original image, and thus the quality of the generated image increases. Experiments demonstrate that the presented method can easily achieve a considerable increase in the peak signal-to-noise ratio and the structural similarity index compared with other methods.

Range query processing is a common edge computing and service in the Internet of things, which can extract user-interest information from distributed edge devices. How to design lightweight privacy-preserving range query processing methods remains a challenging task. Existing secure range query approaches suffer from both high communication cost and long response time, which makes them unsuitable for edge computing over resource-constrained edge devices. In this article, we propose two privacy-aware fuzzy query processing schemes based on fuzzy theory. Linguistic range variables, fuzzy overlap information, and its recovery mechanism are introduced. In addition, two distributed privacy-aware fuzzy range query processing algorithms are devised. Our approaches not only serve for privacy protection, but also aim to provide other optimal performances in terms of reliability, energy efficiency, and real-time response. Theoretical analysis and experimental evaluations based on real-world datasets validated our motivation.

Drivers always confront parking difficulties when driving on urban roads, especially in crowded downtown or beauty spots. Some of the existing literatures concentrate on multi-consideration optimization for parking decision by collecting the nearby real-time parking-related data. Others provide online parking navigation services through outsourced storage and cloud computing. Massive (raw) data transmission and complex processing are always involved in the existing methods, which results in undesired QoS such as real-time performance and privacy protection. In this paper, we propose a fuzzy skyline parking recommendation scheme for real-time parking recommendation based on roadside traffic facilities. Linguistic parking information instead of raw parking-related data is used in fuzzy skyline fusion. We evaluated our solution with real-world data sets collected from edge parking facilities in Wulin downtown, Hangzhou city, China. The evaluation results show that our approaches achieve an average accuracy of parking recommendation over 91%, low data transmission, and quick response time with privacy protection.

With the rapid development of social networks and the massive popularity of intelligent mobile terminals, network anomaly detection is becoming increasingly important. In daily work and life, edge nodes store a large number of network local connection data and audit data, which can be used to analyze network abnormal behavior. With the increasingly close network communication, the amount of network connection and other related data collected by each network terminal is increasing. Machine learning has become a classification method to analyze the features of big data in the network. Face to the problems of excessive data and long response time for network anomaly detection, we propose a trust-based Federated learning anomaly detection algorithm. We use the edge nodes to train the local data model, and upload the machine learning parameters to the central node. Meanwhile, according to the performance of edge nodes training, we set different weights to match the processing capacity of each terminal which will obtain faster convergence speed and better attack classification accuracy. The user’s private information will only be processed locally and will not be uploaded to the central server, which can reduce the risk of information disclosure. Finally, we compare the basic federated learning model and TFCNN algorithm on KDD Cup 99 dataset and MNIST dataset. The experimental results show that the TFCNN algorithm can improve accuracy and communication efficiency.

Event detection is a central task for distributed sensor systems and detecting forthcoming events in a timely manner is the main way of minimizing their possibly damaging effects. The state-of-the-art methods for event description and detection always rely on using crisp raw sensory data, which requires huge data transmission as well as is time-consuming. However, even a centralized processing manner cannot ensure accurate event decision due to the imprecision and uncertainty of raw sensor readings. In many cases, users do not care about the raw sensory data or the data format used for in-network processing, but instead they are concerned with the semantic event information, such as “how serious is it?” and “where will it occur?” In addition, the main technique employed by the existing solution for detecting problems is collaboration with neighbors, which requires massive data exchange between neighbors that is highly intensive in terms of wireless communication . In this paper, we introduce an energy-efficient, reliable semantic event information extraction framework using fuzzy sets . Linguistic event variables instead of raw sensor data are used for event information transmission and fusion, and fuzzy method-based semantic event information filtering and fusion algorithms are proposed. Extensive evaluations based on both real-life and synthetic data sets demonstrated that our framework only incurs a small communication cost and it returns interpretable event information with guaranteed accuracy.

A personally semantic place is a space that is frequently visited by an individual user and carries important semantic meanings (e.g. home, work, etc.) to the user. Many location-aware applications could be greatly enhanced by the ability of automatic discovery of personally semantic places. The discovery of a user's personally semantic places involves obtaining the physical locations and semantic meanings of these places. In this paper, we propose approaches to address both of the problems. For the physical place extraction problem, a hierarchical clustering algorithm is proposed to firstly extract visit points from the GPS trajectories, and then clusters these visit points to form physical places. For the semantic place recognition problem, the temporal, spatial and sequential features in which the places have been visited are explored to categorize them into pre-defined types. An extensive set of experiments conducted based on a dataset of real-world GPS trajectories has demonstrated the effectiveness of the proposed approaches.

Join processing in wireless sensor networks is a challenging problem. Current solutions are not involved in the join operation among tuples of the latest sampling periods. In this article, we proposed a continuous Single attribute Join Queries within latest sampling Periods (SJQP) for wireless sensor networks. The main idea of our filter-based framework is to discard non-matching tuples, and our scheme can guarantee the result is correct independent of the filters. Experiments based on real-world sensor data show that our method performs close to a theoretical optimum and consistently outperforms the centralized join algorithm.

Fuzzy method provides tools to handle imprecise and uncertain data in relevant, robust and interpretable ways. In this paper, we highlight the multi-microenvironments in sensor network and propose a novel event-driven approximate top-k query solution based on fuzzy method. Firstly, membership function of fuzzy method is introduced to describe the global (potential) event confidence in multi-microenvironments sensor network. Subsequently, non-uniform membership degree sub ranges based linguistic labels instead of numeric values are used for in-network event fusion. Also two event-driven approximate top-k query algorithms are devised. Extensive simulations based on real and synthetic data show that our solution reduces the transmission data significantly and the query results are more interpretable with quality guarantees.