检索结果-内蒙古大学图书馆

Lightweight Task-Oriented Semantic Communication Empowered by Large-Scale AI Models

IEEE Transactions on Vehicular technology 2025年

作者： Liu, Chuanhong Guo, Caili Yang, Yang Chen, Mingzhe Quek, Tony Q. S. Beijing University of Posts and Telecommunications Beijing Key Laboratory of Network System Architecture and Convergence School of Information and Communication Engineering Beijing100876 China Beijing University of Posts and Telecommunications Beijing Laboratory of Advanced Information Networks School of Information and Communication Engineering Beijing100876 China University of Miami Department of Electrical and Computer Engineering Institute for Data Science and Computing Coral GablesFL United States Singapore University of Technology and Design Dept. of Information Systems Technology and Design 487372 Singapore

Recent studies have focused on leveraging large-scale artificial intelligence (LAI) models to improve semantic representation and compression capabilities. However, the substantial computational demands of LAI models pose significant challenges for real-time communication scenarios. To address this, this paper proposes utilizing knowledge distillation (KD) techniques to extract and condense knowledge from LAI models, effectively reducing model complexity and computation latency. Nevertheless, the inherent complexity of LAI models leads to prolonged inference times during distillation, while their lack of channel awareness compromises the distillation performance. These limitations make standard KD methods unsuitable for task-oriented semantic communication scenarios. To address these issues, we propose a fast distillation method featuring a pre-stored compression mechanism that eliminates the need for repetitive inference, significantly improving efficiency. Furthermore, a channel adaptive module is incorporated to dynamically adjust the transmitted semantic information based on varying channel conditions, enhancing communication reliability and adaptability. In addition, an information bottleneck-based loss function is derived to guide the fast distillation process. Simulation results verify that the proposed scheme outperform baselines in term of task accuracy, model size, computation latency, and training data requirements. © 1967-2012 IEEE.

关键词： NP-hard

来源：评论

学校读者我要写书评

暂无评论

A Comparative Study of Neural Network Techniques for Automatic Software Vulnerability Detection

A Comparative Study of Neural Network Techniques for Automat...

引用

Theoretical Aspects of Software Engineering (TASE)

作者： Gaigai Tang Lianxiao Meng Huiqiang Wang Shuangyin Ren Qiang Wang Lin Yang Weipeng Cao School of computer science and technology Harbin Engineering University Harbin China National Key Laboratory of Science and Technology on Information System Security Institute of System Engineering Chinese Academy of Military Science Beijing China College of Computer Science and Software Engineering Shenzhen University Shenzhen China

Software vulnerabilities are usually caused by design flaws or implementation errors, which could be exploited to cause damage to the security of the system. At present, the most commonly used method for detecting software vulnerabilities is static analysis. Most of the related technologies work based on rules or code similarity (source code level) and rely on manually defined vulnerability features. However, these rules and vulnerability features are difficult to be defined and designed accurately, which makes static analysis face many challenges in practical applications. To alleviate this problem, some researchers have proposed to use neural networks that have the ability of automatic feature extraction to improve the intelligence of detection. However, there are many types of neural networks, and different data preprocessing methods will have a significant impact on model performance. It is a great challenge for engineers and researchers to choose a proper neural network and data preprocessing method for a given problem. To solve this problem, we have conducted extensive experiments to test the performance of the two most typical neural networks (i.e., Bi-LSTM and RVFL) with the two most classical data preprocessing methods (i.e., the vector representation and the program symbolization methods) on software vulnerability detection problems and obtained a series of interesting research conclusions, which can provide valuable guidelines for researchers and engineers. Specifically, we found that 1) the training speed of RVFL is always faster than Bi-LSTM, but the prediction accuracy of Bi-LSTM model is higher than RVFL; 2) using doc2vec for vector representation can make the model have faster training speed and generalization ability than using word2vec; and 3) multi-level symbolization is helpful to improve the precision of neural network models.

关键词： Training Neural networks Data preprocessing Static analysis Predictive models Feature extraction Software

来源：评论

学校读者我要写书评

暂无评论

Improvements of the CMFD acceleration capability of OpenMOC

Improvements of the CMFD acceleration capability of OpenMOC

引用

2019 International Conference on Mathematics and Computational Methods Applied to Nuclear science and Engineering, M and C 2019

作者： Wu, Wenbin Giudicelli, Guillaume Smith, Kord Forget, Benoit Yao, Dong Luo, Qi Department of Nuclear Science and Engineering Massachusetts Institute of Technology 77 Massachusetts Ave CambridgeMA02139 United States Science and Technology on Reactor System Design Technology Laboratory Nuclear Power Institute of China NO. 328 Section 1 Changshun Ave Shuangliu District Chengdu Sichuan Province610213 China

ISBN: (纸本)9780894487699

Due to its computational efficiency and geometrical flexibility, the Method of Characteristics (MOC) has been widely used for light water reactor lattice physics analysis. Usually acceleration methods are necessary for MOC to achieve acceptable convergence on practical reactor physics problems. Among them, Coarse Mesh Finite Difference (CMFD) is very popular and can drastically reduce the number of transport iterations. In OpenMOC, CMFD acceleration was implemented but had the limitation of supporting only a uniform CMFD mesh, which would often lead to splitting MOC source regions, thus creating an unnecessary increase in computation and memory use. In this study, CMFD acceleration with a nonuniform Cartesian mesh is implemented into OpenMOC. We also propose a quadratic fit based CMFD prolongation method in the axial direction to further improve the acceleration when multiple MOC source regions are contained in one CMFD coarse cell. Numerical results are presented to demonstrate the improvement of the CMFD acceleration capability in OpenMOC in terms of both efficiency and stability. © 2019 American Nuclear Society. All rights reserved.

关键词： Acceleration

来源：评论

学校读者我要写书评

暂无评论

Stability and energetics of two-dimensional surface crystals in liquid AuSi thin films and nanoscale droplets

引用

Physical Review Materials 2021年第7期5卷 074003-074003页

作者： Hailong Wang Mengyuan Wang Alireza Shahabi Alain Karma Moneesh Upmanyu CAS Key Laboratory of Mechanical Behavior and Design of Materials Department of Modern Mechanics CAS Center for Excellence in Complex System Mechanics University of Science and Technology of China Hefei Anhui 230027 China Group for Simulation and Theory of Atomic-Scale Material Phenomena (stAMP) Department of Mechanical and Industrial Engineering Northeastern University Boston Massachusetts 02115 USA Department of Physics Northeastern University Boston Massachusetts 02115 USA

We employ atomic-scale computational frameworks to study the surface crystallization in AuSi films and droplets as a function of composition, temperature, and size. Above the melting point, the surfaces of both thin films and droplets undergo a first-order transition from a two-dimensional (2D) Au2Si crystalline phase to a laterally disordered yet stratified layer. The thin film surfaces exhibit an effective surface tension that increases with temperature and decreases with Si concentration, while for droplets in the size range 10–30 nm, the bulk Laplace pressure alters the surface segregation as it occurs with respect to a strained bulk. Above the transition, the size effect on the surface tension due to the stratified surface layer is small, while the crystalline surface layer below the transition is strained and composed of 2D crystallites separated by extended grain boundary scars that lead to large fluctuations in its energetics. As a specific application, all-atom simulations of AuSi droplets on Si(111) substrate subject to Si surface flux show that the supersaturation dependent surface tension destabilizes the contact line via formation of a precursor wetting film on the solid-vapor interface and has ramifications for size selection during droplet-catalyzed routes for nanowire growth. Our study sheds light on the interplay between stability and energetics of surfaces in this unique class of binary alloys and offers pathways for exploiting their surface structure for varied applications such as catalytic nanocrystal growth, dealloying, and polymer crystallization.

关键词： Crystal forms Growth Interface & surface thermodynamics Nanocrystals Structural phase transition 2-dimensional systems Surfaces Molecular dynamics

来源：评论

学校读者我要写书评

暂无评论

design of High Uniformity and High Energy Efficiency Illumination Lens Based on Hypothetical Point Source Ensemble

引用

Guangxue Xuebao/Acta Optica Sinica 2025年第7期45卷

作者： Li, Ziheng Huo, Furong Xue, Changxi Yang, Chao Dong, Bo School of Optoelectronic Engineering Changchun University of Science and Technology Jilin Changchun130022 China Key Laboratory of Advanced Optical System Design and Manufacturing Technology of Universities of Jilin Province Changchun University of Science and Technology Jilin Changchun130022 China

Objective With the rapid advancement of modern science and technology, illumination systems have become an indispensable part of daily life. In fields such as road, medical, and industrial applications, a well-designed illumination system can significantly enhance safety. Therefore, designing an illumination system with a distributed light pattern that meets human needs has become a key focus in the lighting industry. Light emitting diodes (LEDs), with their advantages of environmental friendliness, energy efficiency, and high performance, have gradually replaced traditional light sources and are now the dominant light solution in the market. However, LEDs often require secondary optical design to achieve more uniform and efficient illumination, meeting the specific lighting needs of different environments and improving overall lighting quality and experience. In this paper, we propose a new method for designing a uniform lighting lens in the context of illumination engineering. Methods In light of the luminous characteristics of LED light sources, a novel hypothesis is proposed for designing extended LED light sources, treating the extended light source as a collection of infinite point sources. The divergence halfangle limit for each small point source is strictly defined, and a uniform internal freeform surface is designed to serve as a secondary light source. The external freeform surface then maps the uniform secondary source to the target surface, a method called the hypothetical point source ensemble (HPE). Based on the inner freeform surface, the outer surface is designed by analyzing the point-to-point energy mapping from the light source to the illuminated plane using the source-target mapping method. We derive the corresponding theoretical model and a new formula for calculating the emitted light angle. The point convergence of the outer freeform surface is calculated using Snell’s law in vector form and the tangent iteration method. The discrete point cl

关键词： Optical design

来源：评论

学校读者我要写书评

暂无评论

A novel TSK fuzzy system incorporating multiview collaborative transfer learning for personalized epileptic EEG detection

arXiv

引用

arXiv 2021年

作者： Li, Andong Deng, Zhaohong Lou, Qiongdan Choi, Kup-Sze Shen, Hongbin Wang, Shitong School of Artificial Intelligence and Computer Science Jiangnan University Jiangsu Key Laboratory of Digital Design and Software Technology Wuxi214122 China Centre for Smart Health Hong Kong Polytechnic University Institute of Image Processing and Pattern Recognition Shanghai Jiao Tong University Shanghai200240 China Key Laboratory of System Control and Information Processing Ministry of Education of China Shanghai200240 China

—In clinical practice, electroencephalography (EEG) plays an important role in the diagnosis of epilepsy. EEG-based computer-aided diagnosis of epilepsy can greatly improve the accuracy of epilepsy detection while reducing the workload of physicians. However, there are many challenges in practical applications for personalized epileptic EEG detection (i.e., training of detection model for a specific person), including the difficulty in extracting effective features from one single view, the undesirable but common scenario of lacking sufficient training data in practice, and the no guarantee of identically distributed training and test data. To solve these problems, we propose a TSK fuzzy system-based epilepsy detection algorithm that integrates multi-view collaborative transfer learning. To address the challenge due to the limitation of single-view features, multi-view learning ensures the diversity of features by extracting them from different views. The lack of training data for building a personalized detection model is tackled by leveraging the knowledge from the source domain (reference scene) to enhance the performance of the target domain (current scene of interest), where mismatch of data distributions between the two domains is resolved with adaption technique based on maximum mean discrepancy. Notably, the transfer learning and multi-view feature extraction are performed at the same time. Furthermore, the fuzzy rules of the TSK fuzzy system equip the model with strong fuzzy logic inference capability. Hence, the proposed method has the potential to detect epileptic EEG signals effectively, which is demonstrated with the positive results from a large number of experiments on the CHB-MIT dataset. © 2021, CC BY-NC-ND.

关键词： Electrophysiology

来源：评论

学校读者我要写书评

暂无评论

The application of Deep Reinforcement Learning in Coordinated Control of Nuclear reactors

引用

Journal of Physics: Conference Series 2021年第1期2113卷

作者： Jing Li Yanyang Liu Xianguo Qing Kai Xiao Ying Zhang Pengcheng Yang Yue (Marco) Yang Science and Technology on Reactor System Design Technology Laboratory Nuclear Power Institute of China Chengdu 610213 China Department of Engineering The University of Sydney Sydney 2006 Australia

The nuclear reactor control system plays a crucial role in the operation of nuclear power plants. The coordinated control of power control and steam generator level control has become one of the most important control problems in these systems. In this paper, we propose a mathematical model of the coordinated control system, and then transform it into a reinforcement learning model and develop a deep reinforcement learning control algorithm so-called DDPG algorithm to solve the problem. Through simulation experiments, our proposed algorithm has shown an extremely remarkable control performance.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Energy recovery characteristics analysis of permanent magnet electrodynamic wheel Maglev vehicle system based on heat-electricity energy conversion strategy under static suspension condition

引用

Energy 2025年 330卷

作者： Le Liang Zhengyan Li Baojian Zhang Peng Pang Zigang Deng National Laboratory of Rail Transit Southwest Jiaotong University Chengdu Sichuan 610031 China Key Laboratory of Magnetic Suspension Technology and Maglev Vehicle Ministry of Education Southwest Jiaotong University Chengdu Sichuan 610031 China School of Information Science and Technology Southwest Jiaotong University Chengdu 610031 China State Key Laboratory of Rail Transit Vehicle System Southwest Jiaotong University Chengdu Sichuan 610031 China School of Electrical Engineering Southwest Jiaotong University Chengdu Sichuan 610031 China

In this paper, a heat-electricity energy conversion strategy based on thermoelectric theory in permanent magnet electrodynamic wheel (PMEDW) magnetic levitation (Maglev) vehicle system is proposed and tested for the first time. The PMEDW-Maglev vehicle is based on electromagnetic induction interaction principle between PMEDW and conductor plate, to generate levitation force and propulsion force. Hence, the PMEDW-Maglev vehicle is a new concept with levitation-propulsion multifunctional integrated design, to supply the faster, greener, smarter highway transportation mode. But, the temperature rise in conductor plate may cause the levitation failure and reduce the system operation stability. Hence, considering energy effective conservation and utilization, the thermoelectric theory based heat-electricity energy conversion strategy is proposed in this paper, to utilize the heat generated by temperature rise and improve system operation performance. The operation performance of PMEDW-Maglev prototype vehicle with the heat-electricity energy conversion strategy is tested by the experiment. It can be found from the experimental results that, the proposed heat-electricity energy conversion strategy can operate as expected. The levitation force and temperature rise cooling ratio in conductor plate can be improved by more than 47 % and 12 %, respectively, besides, the heat can be converted into electricity energy and stored in battery unit (more than 280 J). The significance of this work is that the proposed heat-electricity energy conversion strategy can offer a performance enhancement and energy self consistency design method for the future PMEDW-Maglev vehicle system.

关键词： Energy conservation Energy system self consistency Heat-electricity energy conversion PMEDW- Maglev vehicle

来源：评论

学校读者我要写书评

暂无评论

Neutron physical feasibility of small modular design of dual fluid reactor 27

Neutron physical feasibility of small modular design of dual...

引用

27th International Conference on Nuclear Engineering: Nuclear Power Saves the World!, ICONE 2019

作者： Wang, Xiang Zhang, Qian He, Xun Du, Zhuoqi Seidl, Marcus Macian-Juan, Rafael Czerski, Konrad Dabrowski, Mariusz College of Nuclear Science and Technology Harbin Engineering University Harbin Heilongjiang China Science and Technology on Reactor System Design Technology Laboratory Nuclear Power Institute of China Chengdu Sichuan China Institute of Nuclear Engineering Technical University of Munich Garching Bayern Germany Faculty of Mathematics and Physics University of Szczecin Szczecin Poland

ISBN: (纸本)9784888983051

The Dual Fluid reactor (DFR) is a fast reactor concept proposed by the Institute of Solid-state- and Nuclear physics (IFK) in Berlin. The design of DFR aims to combine the Gen-IV Molten Salt reactor (MSR) and the Liquid-Metal Cooled reactor (SFR, LFR), which means that the molten-salt fuel is no more used as coolant while the heat is removed in a separate loop by liquid lead, to improve these two concepts. Without control rods, DFR is given hope to react and take steps to the change of operation conditions automatically by its inertial negative temperature feedback. Since fuel and coolant are both liquid, the circulation speed of each liquid is considered capable to adjust operational parameters for DFR. The original design of a 3000MWth (Wang, 2017) and a downscaled version of 500MWth (He, 2016) were analyzed in the aspects of reactor physics and thermal-hydraulics. For the practical purpose, this paper has proposed a small modular version of the DFR, which has power output of 2MWth. The reactor with such power level is expected in various applications. Therefore, it is necessary to perform the more explicit analysis on the design than the previous ones. The analysis focuses on the neutronic calculations with the default fuel salt of U-Pu mixture under steady state conditions. Copyright © 2018-2019 by JSME

关键词： Coolants

来源：评论

学校读者我要写书评

暂无评论

Research on active disturbance rejection control method for electro-hydraulic servo loading system with considering clearance disturbance 8

Research on active disturbance rejection control method for ...

引用

8th IEEE International Conference on Fluid Power and Mechatronics, FPM 2019

作者： Chen, Qiang Yang, Fang Liu, Dan-Dan Xue, Yu-Jun Henan University of Science Technology School of Mechatronics Engineering Luoyang China Henan University of Science Technology Henan Key Laboratory of Mechanical Design and Transmission System Luoyang China Luoyang LYC Rearing Co. Ltd Luoyang China

ISBN: (纸本)9781728103112

In the test of high-speed rail bearings, the electro- hydraulic servo-loading system is commonly used to simulate the actual alternating loading for the bearing. Due to the existence of the loading mechanism clearance and the transmission shaft system matching clearance, the loading system has a certain nonlinear displacement disturbance. Combined with the nonlinear characteristics of the electro-hydraulic servo system, the dynamic response characteristics and steady-state control accuracy of the force loading are seriously affected. To this end, the Active Disturbance Rejection control (ADRC) method is used to identify and compensate the gap disturbances using the feedback information of the extended state observer, and the tracking differentiator and feedback control rate are designed to improve the system response speed and suppress overshoot. The control method is applied to the control model of high-speed rail bearing electro-hydraulic servo-loading system. The simulation results show that the control method can achieve good control effect, and the anti-disturbance capability, response speed and steady-state performance of the system are significantly improved. © 2019 IEEE.

关键词： Disturbance rejection

来源：评论

学校读者我要写书评

暂无评论

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案：

请选择收藏分类：

通借通还

建议与咨询 留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

时间限定

文献类型

馆藏选择

核心期刊

语言

文献类型

帮助

文字说明：

检索规则说明：

检索范例：

分类表

所选分类

限定检索结果

文献类型

馆藏范围

日期分布

学科分类号

主题

机构

作者

语言

请选择保存的检索档案： 新增检索档案 确定 取消

请选择收藏分类： 新增自定义分类 确定 取消

通借通还

建议与咨询留下您的常用邮箱和电话号码，以便我们向您反馈解决方案和替代方法

请选择保存的检索档案：

请选择收藏分类：