- Parallel Computing and Optimization Techniques
- Advanced Neural Network Applications
- Embedded Systems Design Techniques
- Interconnection Networks and Systems
- Graph Theory and Algorithms
- Advanced Image and Video Retrieval Techniques
- CCD and CMOS Imaging Sensors
- Cloud Computing and Resource Management
- Advanced Algorithms and Applications
- Advanced Computational Techniques and Applications
- Adversarial Robustness in Machine Learning
- Advanced Queuing Theory Analysis
- Advanced Data Storage Technologies
- Advanced Vision and Imaging
- Formal Methods in Verification
- Network Security and Intrusion Detection
- Matrix Theory and Algorithms
- Petri Nets in System Modeling
- Advancements in Semiconductor Devices and Circuit Design
- Sensorless Control of Electric Motors
- VLSI and FPGA Design Techniques
- Advanced Image Processing Techniques
- Advanced Memory and Neural Computing
- Advanced Graph Neural Networks
- IoT and Edge/Fog Computing
Hubei University of Technology
2025
National University of Singapore
2008-2025
Stomatology Hospital
2025
Sichuan University
2020-2025
State Key Laboratory of Oral Diseases
2025
East China Jiaotong University
2024
Wuhan Donghu University
2023-2024
Shanghai University of International Business and Economics
2018-2024
Ministry of Public Security of the People's Republic of China
2013-2024
China University of Geosciences
2024
The goal of video summarization is to distill a raw into more compact form without losing much semantic information. However, previous methods mainly consider the diversity and representation interestingness obtained summary, they seldom pay sufficient attention information resulting frame set, especially long temporal range semantics. To explicitly address this issue, we propose novel technique which able extract most semantically relevant segments (i.e., valid for term duration) assemble...
The efficacy and effectiveness of Convolutional Neural Networks (CNNs) have been proven in a wide range machine learning applications. However, the high computational complexity CNNs presents critical challenge towards their broader adoption real-time power-efficient scenarios. FPGAs are poised to take significant role for high-performance energy-efficient computation both mobile (e.g., UAVs, self-driving cars, IoT devices) cloud computing domains. implementing an effective CNN system onto...
High quality AI solutions require joint optimization of algorithms and their hardware implementations. In this work, we are the first to propose a fully simultaneous, Efficient Differentiable DNN (deep neural network) architecture implementation co-search (EDD) methodology. We formulate problem by fusing search variables into one solution space, maximize both algorithm accuracy quality. The formulation is differentiable with respect fused variables, so that gradient descent can be applied...
Abstract Pre-trained Transformer-based models have achieved state-of-the-art performance for various Natural Language Processing (NLP) tasks. However, these often billions of parameters, and thus are too resource- hungry computation-intensive to suit low- capability devices or applications with strict latency requirements. One potential remedy this is model compression, which has attracted considerable research attention. Here, we summarize the in compressing Transformers, focusing on...
FPGA has been an emerging computing infrastructure in datacenters benefiting from features of fine-grained parallelism, energy efficiency, and reconfigurability. Meanwhile, graph processing attracted tremendous interest data analytics, its performance is increasing demand with the rapid growth data. Many works have proposed to tackle challenges designing efficient FPGA-based accelerators for processing. However, largely overlooked programmability still requires hardware design expertise...
Over the past few years, chaotic image encryption has gained extensive attention. Nevertheless, current studies on still possess certain constraints. To break these constraints, we initially created a two-dimensional enhanced logistic modular map (2D-ELMM) and subsequently devised scheme based vector-level operations 2D-ELMM (CIES-DVEM). In contrast to some recent schemes, CIES-DVEM features remarkable advantages in several aspects. Firstly, is not only simpler structure, but its performance...
This paper presents the first parameterized SPICE-compatible compact model of a graphene nano-ribbon field-effect transistor (GNRFET) with doped reservoirs, also known as MOS-type GNRFET. The current and charge models closely match numerical TCAD simulations. In addition, process variation in dimension, line edge roughness, doping level reservoirs are accurately modeled. Our provides means to analyze delay power graphene-based circuits under variation, offers design fabrication insights for...
Performance of object detection models has been growing rapidly on two major fronts, model accuracy and efficiency. However, in order to map deep neural network (DNN) based edge devices, one typically needs compress such significantly, thus compromising the accuracy. In this paper, we propose a novel GPU friendly module for multi-scale feature interaction by exploiting missing combinatorial connections between various scales existing state-of-the-art methods. Additionally, transfer learning...
Open government data (OGD) has attracted widespread attention and been widely carried out on a global scale. With further promotion, OGD performance becomes hot topic meaningful enough for in-depth exploration. This research focuses the influential factors generation mechanisms of performance. Based resource-based theory institutional theory, this paper constructs model from multiple dimensions internal resources external pressures. Subsequently, 122 cities in China that have constructed...
This paper addresses two aspects of low-power design for FPGA circuits. First, we present an RT-level power estimator FPGAs with consideration wire length. The closely reflects both dynamic and static contributed by various components in 0.1 /spl mu/m technology. estimation error is 16.2% on average. Second, a low high level synthesis system, named LOPASS, designs. It includes algorithms consumption reduction: (i) simulated annealing engine that carries out resource selection function unit...
As the trends move towards data outsourcing and cloud computing, efficiency of distributed centers increases in importance. Cloud-based services such as Amazon's EC2 rely on virtual machines (VMs) to host MapReduce clusters for large processing. However, current VM scheduling does not provide adequate support workloads, resulting degraded overall performance. For example, when multiple run a single physical machine, existing VMMscheduler guarantee fairness across clusters.
In this paper we aim to understand the types of applications for which cloud computing is economically tenable, i.e., cost savings associated with placement outweigh any deployment costs.
High level synthesis (HLS) is gaining wider acceptance for hardware design due to its higher productivity and better space exploration features. In recent years, HLS techniques flows have also advanced significantly, as a result, many new FPGA designs are developed with HLS. However, despite studies using HLS, the size complexity of such applications remain generally small, it not well understood how optimize large, complex reference code. Typical benchmark contain somewhere between 100 1400...
What is known and objective Drug-drug interactions (DDI) are frequent causes of adverse clinical drug reactions. Efforts have been directed at the early stage to achieve accurate identification DDI for safety assessments, including development in silico predictive methods. In particular, similarity-based methods developed assess with good accuracies, machine learning employed further extend range approaches. However, performance a method lower than expectations partly because use less...
This paper presents an accurate analytical compact model for Schottky-barrier-type graphene nanoribbon field-effect transistors (SB-GNRFETs). is a physics-based the current-voltage (I-V) characteristics of SB-GNRFETs. The proposed considers various design parameters and process variation effects, including graphene-specific line-edge roughness, which allows thorough exploration evaluation SB-GNRFET circuits. We develop approximations SB tunneling, channel charge, current, provide results...
Unsupervised node embedding methods (e.g., DeepWalk, LINE, and node2vec) have attracted growing interests given their simplicity effectiveness. However, although these been proved effective in a variety of applications, none the existing work has analyzed robustness them. This could be very risky if are attacked by an adversarial party. In this paper, we take task link prediction as example, which is one most fundamental problems for graph analysis, introduce data positioning attack to...
Quantization has been proven to be an effective method for reducing the computing and/or storage cost of DNNs. However, trade-off between quantization bitwidth and final accuracy is complex non-convex, which makes it difficult optimized directly. Minimizing direct loss (DQL) coefficient data local optimization method, but previous works often neglect accurate control DQL, resulting in a higher DNN model accuracy. In this paper, we propose novel metric called Vector Loss. Based on new metric,...
Control and communication technologies are key building blocks of cyber-physical systems (CPSes) that can improve the efficiency physical processes. However, they also make a CPS vulnerable to cyberattacks cause disruptions or even severe damage. This article focuses on one particular type cyberattack, namely time delay attack (TDA), which exploits vulnerabilities in channels potentially serious harm system. Much work proposed for TDA detection is tested offline only under strong...
To solve the problem that fault features are difficult to extract and time-frequency cannot fully represent state information, a novel method is proposed in this paper based on whale optimization algorithm (WOA) kernel extreme learning machine (KELM). First, vibration signals processed by ensemble empirical mode decomposition sample entropy obtain feature vectors. Based this, KELM model for diagnosis established. Then, penalty factor parameters optimized WOA improve stability classification...
The increasing demand for efficient summarization tools in resource-constrained environments highlights the need effective solutions. While large language models (LLMs) deliver superior quality, their high computational resource requirements limit practical use applications. In contrast, small (SLMs) present a more accessible alternative, capable of real-time on edge devices. However, capabilities and comparative performance against LLMs remain underexplored. This paper addresses this gap by...
The effectiveness of in-memory dynamic graph storage (DGS) for supporting concurrent read and write queries is crucial real-time analytics updates. Various methods have been proposed, example, LLAMA, Aspen, LiveGraph, Teseo, Sortledton. These approaches differ significantly in their support operations, space overhead, concurrency control. However, there has no systematic study to explore the trade-offs among these dimensions. In this paper, we evaluate individual techniques identify...