检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Zhang, Songge Cheng, Guoliang Huang, Xinyu Li, Zuguang Wu, Wen Song, Lingyang Shen, Xuemin School of Electronic and Computer Engineering Peking University Shenzhen518000 China Frontier Research Center Pengcheng Laboratory Shenzhen518055 China State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Peking University Beijing100871 China School of Electronic and Computer Engineering Peking University Shenzhen Graduate School Shenzhen518055 China Department of Electrical and Computer Engineering University of Waterloo WaterlooN2L 3G1 Canada

Fine-tuning is the process of adapting the pre-trained large language models (LLMs) for downstream tasks. Due to substantial parameters, fine-tuning LLMs on mobile devices demands considerable memory resources, and suffers from high communication overhead and long fine-tuning delay. In this paper, we propose an efficient LLM fine-tuning scheme in wireless networks, named Split Fine-Tuning (SFT), which can accommodate LLM fine-tuning on mobile devices. Specifically, an LLM is split into a server-side part on the edge server and a device-side part on the mobile device to satisfy the device-side memory constraint. All devices share a server-side model and perform parallel fine-tuning to reduce fine-tuning delay. In addition, to reduce significant communication overhead incurred by data exchange between devices and the edge server, we propose a data compression scheme by jointly leveraging sparsification, stochastic quantization, and lossless encoding methods. Furthermore, we formulate a fine-tuning delay minimization problem under accuracy and memory constraints, taking device heterogeneity and channel dynamics into account. To solve the problem, the nonlinear mixed-integer problem is decoupled into two subproblems in different timescales. The two-timescale resource management algorithm is proposed to jointly optimize the compression rate and transformer block allocation in the large timescale using the augmented Lagrangian method, and determine spectrum resource allocation in the small timescale via sequential quadratic programming. Extensive simulation results demonstrate that the proposed scheme can reduce the fine-tuning delay by up to 80.2% and communication overhead by 93.6% compared to state-of-the-art benchmarks, while satisfying device-side memory and model accuracy constraints. Copyright © 2025, The Authors. All rights reserved.

关键词： Quadratic programming

来源：评论

学校读者我要写书评

暂无评论

Beam Pattern Modulation Embedded Hybrid Transceiver Optimization for Integrated Sensing and communication

引用

IEEE Transactions on Wireless communications 2025年

作者： Liu, Boxun Gao, Shijian Yang, Zonghui Cheng, Xiang Yang, Liuqing Peking University State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Beijing100871 China Internet of Things Thrust Guangzhou511400 China Internet of Things Thrust Intelligent Transportation Thrust Guangzhou511400 China The Hong Kong University of Science and Technology Department of Electronic and Computer Engineering Department of Civil and Environmental Engineering SAR Hong Kong Hong Kong

Integrated sensing and communication (ISAC) emerges as a promising technology for 6G, particularly in the millimeter-wave (mmWave) band. However, the widely utilized hybrid architecture in mmWave systems compromises multiplexing gain due to the constraints of limited radio-frequency (RF) chains. Moreover, additional sensing functionalities exacerbate the impairment of spectrum efficiency (SE). In this paper, we present an optimized beam pattern modulation-embedded ISAC (BPM-ISAC) transceiver design, which spares one RF chain for sensing and uses the remaining ones for communication. To compensate for the reduced SE, index modulation across communication beams is applied. We formulate an optimization problem aimed at minimizing the mean squared error (MSE) of the sensing beampattern, subject to a symbol MSE constraint. This problem is then solved by sequentially optimizing the analog and digital parts. Both the multi-aperture structure (MAS) and the multi-beam structure (MBS) are considered in the analog part. We conduct theoretical analysis on the asymptotic pairwise error probability (APEP) and the Cramér-Rao bound (CRB) of direction of arrival (DoA) estimation. Numerical simulations validate the overall enhanced ISAC performance over existing alternatives. © 2025 IEEE. All rights reserved.

关键词： Mean square error

来源：评论

学校读者我要写书评

暂无评论

A new method for extending rate equation based VCSEL model with multimode spectral characteristic

引用

Science China(Information Sciences) 2018年第12期61卷 155-162页

作者： Anzhong LIANG Chuanchuan YANG Cheng ZHANG Jingjing SONG Hongbin LI State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Engineering and Computer Science Peking University

An accurate VCSEL model is important to evaluate the performance of VCSEL-based high speed short range optical interconnect system. Current VCSEL models show no evidence of the multimode spectral characteristic, which leads to an inaccurate performance estimation of VCSEL-based high speed optical interconnect system especially with large chromatic dispersion. To overcome this problem, we propose a new method to introduce multimode spectral characteristic into current VCSEL models, which extends the spatiotemporal multimode rate equations by calculating the wavelength shift of the VCSEL transverse modes. A VCSEL-multimode fiber(MMF) based short range optical interconnect simulation system is then built using our modified VCSEL model by VPItransmissionMaker. The symbol broadening and the BER performance of 15 Gb/s NRZ signal are tested in both left-tilted differential mode delay(DMD) profile MMF(L-MMF) and right-tilted DMD profile MMF(R-MMF). Simulation results validate our proposed method and indicate that the spectral characteristic is a key component for VCSEL models.

关键词： short-range optical interconnect vertical cavity surface emitting lasers VCSEL multimode rate equations multimode spectral characteristic multimode fiber

来源：评论

学校读者我要写书评

暂无评论

Error correction for short-range optical interconnect using COTS transceivers

引用

Science China(Information Sciences) 2019年第6期62卷 72-78页

作者： Ziyuan ZHENG Chuanchuan YANG Dan ZHAO Ziyu WANG State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Engineering and Computer Science Peking University

Nowadays, quantities of commercial off-the-shelf(COTS) optical transceivers are widely equipped in both exiting and abuilding data centers. As data centers are becoming larger and containing more servers, the method to ensure longer and steady transmission with COTS optical transceivers becomes one of the key issues. Meanwhile, variations in the manufacturing of COTS optical transceivers bring in fluctuation of packet error rate(PER) in communication. In this paper, we present an application layer adaptive forward error correction(AL-AFEC) coding scheme for COTS optical transceivers in optical interconnect to solve the problems mentioned above and test its performance in a 10 Gbps Ethernet(10 GbE)link using 10 GBASE-SR SFP+ modules. Experimental results show that the proposed scheme enables longer transmission distance and the instability introduced by manufactural variations is well settled. Since our scheme requires no extra hardware on optical transceivers, it is a considerable low-cost alternative to improve system reliability and achieve longer transmission distance for COTS optical transceivers.

关键词： short-range optical interconnect vertical cavity surface emitting laser(VCSEL) ethernet application layer adaptive forward error correction commercial off-the-shelf optical transceiver

来源：评论

学校读者我要写书评

暂无评论

Pseudo pilot-aided OFDM system with high spectral efficiency for high performance coaxial transmission

引用

Science China(Information Sciences) 2017年第6期60卷 191-199页

作者： Yue LIU Chuanchuan YANG Hongbin LI State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Engineering and Computer Science Peking University

High performance network over coax(HINOC) has become next-generation broadcasting network(NGB) industry standard of China, as well as International Telecommunication Union(ITU) telecommunication standard, which adopts orthogonal frequency division multiplexing(OFDM) technique and high order modulation formats. As is known, in OFDM system, phase noise destroys the orthogonality of the subcarriers and degrades the performance of the system especially when the data are transmitted in high order modulation formats. In this paper, we propose a high spectral efficiency OFDM system utilizing a novel and almost blind phase noise compensation method mainly used for high performance coaxial transmission. In our system, some subcarriers, which remain to transmit unknown useful data in lower order modulation formats, are selected as pseudo pilots instead of pilots to suppress phase noise effectively and high spectral efficiency can be achieved with the aid of puncturing mechanisms. Simulation results show that the proposed pseudo pilot-aided OFDM system can achieve the performance of the pilot-aided counterpart yet acquiring higher spectral efficiency.

关键词： orthogonal frequency division multiplexing(OFDM) high performance network over coax(HINOC) phase noise pseudo pilots broadcasting

来源：评论

学校读者我要写书评

暂无评论

Deep belief network-hidden Markov model based nonlinear equalizer for VCSEL based optical interconnect

引用

Science China(Information Sciences) 2020年第6期63卷 155-163页

作者： Fukui TIAN Chuanchuan YANG State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Engineering and Computer Science Peking University

The data center has developed rapidly over the past few years, leading to the demand for high speed data transmission. Vertical cavity surface emitting lasers(VCSELs) based optical interconnect is evolving to 100 Gb/s and beyond, which makes nonlinear distortions difficult to be compensated or equalized by conventional equalizers. Moreover, the challenge becomes very complicated for conventional equalizers because of the presence of inter-symbol interference(ISI) together with the nonlinear distortions. So many neural network based DSP algorithms such as artificial neural network(ANN) have been proposed to mitigate the distortions. However, ANN has the limitations that sample's relevant information is not considered,leading to degradation in ANN's performance and high computational complexity. In this paper, in order to maintain an excellent capability of mitigating nonlinear distortions like other neural network based equalizers while considering the sample's relevant information to reduce the computational complexity, we propose a deep belief network-hidden Markov model(DBN-HMM) based nonlinear equalizer which is tested in a PAM-4 modulated VCSEL and multimode fiber(MMF) optical interconnect link experimentally. The BER performance can be greatly improved compared with conventional DSP algorithms. In addition, the computational complexity of DBN-HMM based equalizer can be about 41% lower than that of ANN based method with a similar BER performance.

关键词： short-range optical interconnect vertical cavity surface emitting laser(VCSEL) nonlinear equalizer deep belief network(DBN) hidden Markov model(HMM)

来源：评论

学校读者我要写书评

暂无评论

引用

Frontiers of Information Technology & Electronic Engineering 2017年第6期18卷 796-807页

作者： Zhen-hua YUAN Chen CHEN Xiang CHENG Guo-cheng LV Liu-qing YANG Ye JIN State Key Laboratory of Advanced Optical Communication Systems and Networks Peking University Department of Electrical & Computer Engineering Colorado State University

This article is focused on secure relay beamformer design with a correlated channel model in the relay-eavesdropper network. In this network, a single-antenna source-destination pair transmits secure information with the help of an amplify-and-forward (AF) relay equipped with multiple antennas, and the legitimate and eavesdropping channels are correlated. The relay cannot obtain the instantaneous channel state information (CSI) of the eavesdropper, and has only the knowledge of correlation information between the legitimate and eavesdropping channels. Depending on this information, we derive the conditional distribution of the eavesdropping channel. Two beamformers at the relay are studied for the approximate ergodic secrecy rate:(1) the generalized match-andforward (GMF) beamformer to maximize the legitimate channel rate, and (2) the general-rank beamformer (GRBF).In addition, one lower-bound-maximizing (LBM) beamformer at the relay is discussed for maximizing the lower bound of the ergodic secrecy rate. We find that the GMF beamformer is the optimal rank-one beamformer, that the GRBF is the iteratively optimal beamformer, and that the performance of the LBM beamformer for the ergodic secrecy rate gets close to that of the GRBF for the approximate secrecy rate. It can also be observed that when the relay has lower power or the channel gain of the second hop is low, the performance of the GMF beamformer surpasses that of the GRBF. Numerical results are presented to illustrate the beamformers' performance.

关键词： Physical layer security Relay beamforming Correlated channels Ergodic secrecy rate

来源：评论

学校读者我要写书评

暂无评论

Towards an intelligent photonic system

引用

Science China(Information Sciences) 2020年第6期63卷 80-96页

作者： Weiwen ZOU Bowen MA Shaofu XU Xiuting ZOU Xingjun WANG State Key Laboratory of Advanced Optical Communication Systems and Networks Intelligent Microwave Lightwave Integration Innovation Center (iMLic) Department of Electronic EngineeringShanghai Jiao Tong University State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Engineering and Computer Science Peking University

The emerging intelligence technologies represented by deep learning have broadened their applications to various fields. Beyond the conventional electronics-based processing systems, the convergence of photonics and artificial intelligence(AI) technology enhances the performance and learning ability of AI. In this review, we propose the concept of an intelligent photonic system(IPS), illustrating it as a developing architecture with three different versions. For each version of IPS, we review several representative *** we discuss the challenges towards an IPS and provide some prospects for the future development.

关键词： intelligent photonic system deep learning artificial intelligence optical neural network neuromorphic photonics

来源：评论

学校读者我要写书评

暂无评论

Energy-Efficient Beamforming for Two-Tier Massive MIMO Downlink

引用

China communications 2015年第10期12卷 64-75页

作者： XU Guozhen LIU An JIANG Wei XIANG Haige LUO Wu China Mobile Research Institute & The State Key Laboratory of Advanced Optical Communication Systems and Networks Peking University Department of Electrical and Computer Engineering Hong Kong University of Science and Technology The State Key Laboratory of Advanced Optical Communication Systems and Networks Peking University

Heterogeneous networks(HetNets)consisting of macro cells with very large antenna arrays and a secondary tier of small cells with a few antennas each can well tackle the contradiction of large coverage of the network and high data rate at the hot ***,it is not permissible to assign orthogonal pilot sequences for all the supported users due to the large ***,we propose a pilot reduction scheme based on the heterogeneous system configurations and the unique topology of this *** reusing of pilot sequences causes the presence of the contaminated channel state information(CSI) and results in receivers' Quality of Service(QoS) *** the contaminated CSI,we provide an energy-efficient beamforming based on minimizing the total power consumption while keeping the QoS constraints satisfied and restricting the QoS outage probability below a given *** applying the approach of Bernstein approximation and semi-definite relaxation,we transform the original intractable chance constrained program to a convex problem *** results show that the average power consumption of the proposed beamforming for our pilot reduction scheme is close to that of the perfect CSI *** our scheme will greatly compress the length of pilot sequence especially for those highly densified network with large number of small cells,it will be crucially helpful to put such two-tier massive multiple-input and multiple-output(MIMO) systems into practice.

关键词： massive MIMO small cell Het-Net energy efficiency

来源：评论

学校读者我要写书评

暂无评论

Computing Power in the Sky: Digital Twin-assisted Collaborative Computing with Multi-UAV networks

引用

IEEE Transactions on Vehicular Technology 2025年

作者： Wang, Chao Han, Yu Zhang, Long Jia, Ziye Zhang, Hongliang Hong, Choong Seon Han, Zhu Hebei University of Engineering School of Information and Electrical Engineering Handan056038 China Qilu University of Technology Shandong Academy of Sciences Key Laboratory of Computing Power Network and Information Security Ministry of Education Jinan250014 China Nanjing University of Aeronautics and Astronautics College of Electronic and Information Engineering Nanjing211106 China Southeast University National Mobile Communications Research Laboratory Jiangsu Nanjing211111 China Peking University State Key Laboratory of Advanced Optical Communication Systems and Networks School of Electronics Beijing100871 China Kyung Hee University Department of Computer Science and Engineering Gyeonggi-do Yongin-si17104 Korea Republic of University of Houston Department of Electrical and Computer Engineering HoustonTX77004 United States Kyung Hee University Department of Computer Science and Engineering Seoul446-701 Korea Republic of

The collaboration of computing powers (CPs) among unmanned aerial vehicles (UAVs)-mounted edge servers is essential to handle data-intensive tasks of user equipments (UEs). This paper presents a multi-UAV computing power network (CPN) that orchestrates the CP resources of edge servers to cooperatively process tasks through collaborative computing. Due to the frequent interactions among edge servers and the increased complexities for large-scale task executions, we propose a digital twin (DT)-driven multi-UAV CPN framework, where the aerial DT network is created as a digital replica of the multi-UAV network to assist collaborative computing. A joint optimization problem of the service association, offloading power control, task partitions, CP allocation, and UAV trajectory design is formulated to minimize the long-term average delay of UEs. We use the Lyapunov technique to tackle the long-term energy constraint, forming a tractable optimization problem, which is modeled as a Markov decision process to balance the UAV's deficit queue stability and the UE's task completion delay. A hybrid soft actor-critic based deep reinforcement learning algorithm is developed to optimize the joint design by sampling the hybrid action from the Gumbel and Gaussian distributions. Simulation results verify that the multi-UAV CPN contributes to significant performance gains when supported by DTs. © 2025 IEEE.

关键词： Markov processes

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：