检索结果-内蒙古大学图书馆

IEEE Wireless Communications and networking Conference (IEEE WCNC)

作者： Zhou, Ai Wang, Ying Zhang, Qiuyang Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350303582;9798350303599

Federated learning (FL) is a paradigm that enables model training across various devices while keeping the data localized. However, for battery-powered passive devices in the satellite Internet of Things (IoT), the continuous update and transmission of the local model result in heightened energy consumption on the device side. To address this challenge, an FL framework with partial device participating is proposed. In this framework, the on-board controller strategically selects a subset of devices to upload local model parameters, effectively mitigating the overall energy consumption on the device side. Constrained by transmission power and transmission delay, a resource allocation problem is formulated. This problem jointly optimizes the uploading strategy and transmission power, aiming to minimize the utility function that combines the global model loss and energy consumption over multiple rounds of FL. Simulation results demonstrate that, compared with other benchmark schemes (DDPG, PPO), the proposed algorithm achieves energy efficiency in FL.

关键词： Federated learning satellite Internet of Things deep reinforcement learning resource allocation

来源：评论

学校读者我要写书评

暂无评论

Adaptive Clipping and Distillation Enabled Federated Unlearning

Adaptive Clipping and Distillation Enabled Federated Unlearn...

引用

IEEE International Conference on Web Services (IEEE ICWS)

作者： Xie, Zhiqiang Gao, Zhipeng Lin, Yijing Zhao, Chen Yu, Xinlei Chai, Ze Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350368567;9798350368550

With the advancement of federated crowdsourcing services, the associated privacy concerns have attracted growing attention from both academia and industry. Existing privacy laws impose strict requirements concerning the right to be forgotten for data used in training AI models. In federated crowdsourcing services, the right to be forgotten is guaranteed through federated unlearning. Current federated unlearning solutions encompass a two-step process: first, eliminating model updates associated with the target data to achieve unlearning, followed by retraining among the remaining clients to restore the performance of federated crowdsourcing services. However, this indiscriminate removal of model updates, while safeguarding the privacy of the target data, also greatly undermines the generalization performance of the global model. Moreover, relying on client-side retraining imposes additional economic costs on the federated crowdsourcing service. To tackle the above issues, this paper proposes an efficient federated unlearning framework for federated crowdsourcing services, which is based on adaptive parameter clipping and data-free distillation. We first compute the Fisher information matrix (FIM) to approximate the correlation between the target data and all model parameters, which is utilized to adaptively clip each parameter of the global model. Then, we model the softmax layer of the global model to synthesize pseudo-samples, enabling the retrain process on the crowdsourcing platform for the recovery of generalization performance. We conducted extensive experiments on three datasets, and the results demonstrate that our proposed framework not only possesses outstanding data removal capability but also outperforms the comparison methods in terms of computation time and storage space.

关键词： Federated crowdsourcing services Federated Unlearning Fisher Information Matrix Data-free Knowledge Distillation

来源：评论

学校读者我要写书评

暂无评论

A Novel 3D Non-Stationary Wideband GBSM for AIRS-Assisted Massive MIMO V2V Communication in Viaduct Scenarios 25

A Novel 3D Non-Stationary Wideband GBSM for AIRS-Assisted Ma...

引用

IEEE Wireless Communications and networking Conference (IEEE WCNC)

作者： Peng, Ke Ma, Nan Zhang, Kaiheng Liu, Shaoyi Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350303582;9798350303599

Massive multiple-input multiple-output (MIMO) vehicle-to-vehicle (V2V) communication faces more challenges in viaduct scenarios due to the obstructive effects of ramps and sound barriers compared to normal urban roads. To improve the communication quality, aerial intelligent reflecting surface (AIRS) can be flexibly deployed in this scenario to assist V2V communication. This paper presents a novel three dimensional (3D) non-stationary wideband geometry-based stochastic model (GBSM) designed for enhancing MIMO V2V communication systems utilizing AIRS in viaduct scenarios. Leveraging this framework, the paper derives closed-form expressions for the channel impulse responses (CIR) under spherical wavefront assumption. Furthermore, correlation functions and channel capacities are analytically obtained and several viable phase-shift schemes are tailored for the AIRS units. Simulation results show that the proposed model can accurately capture the characteristics of the AIRS-assisted MIMO V2V channel in viaduct scenarios, and the proposed phase-shift methods can effectively improve the channel statistical characteristics and increase the system capacity.

关键词： 3D wideband GBSM massive MIMO aerial intelligent reflecting surface (AIRS) V2V viaduct scenarios

来源：评论

学校读者我要写书评

暂无评论

Federated Domain Generalization for Network Traffic Prediction via Spatial-Temporal Feature Learning 29

Federated Domain Generalization for Network Traffic Predicti...

引用

29th IEEE Symposium on Computers and Communications (IEEE ISCC)

作者： Chen, Shaoyin Yang, Yang Mei, Jingting Gao, Zhipeng Rui, Lanlan Yu, Peng Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350354249;9798350354232

Network traffic prediction is crucial for network operation and management, forming the basis for utilizing big data in decision support. Traditional deep learning methods require extensive, assumed independently and identically distributed(IID) data. However, with IoT development, privacy protection gains importance, resulting in distributed data collection with varying distributions. This leads to a significant performance drop when applying a well-trained model to a new dataset, causing domain shift. To tackle domain shift from inconsistent data distributions and meet privacy protection needs, this paper proposes a spatial-temporal feature learning method within the federated domain generalization framework for network traffic prediction. The ultimately trained model effectively generalizes to an unseen domain, as confirmed by experimental results.

关键词： federated domain generalization TCN depth-wise separable convolution Transformer dynamic weights network traffic prediction

来源：评论

学校读者我要写书评

暂无评论

Phase Optimization Approaches for RIS-Aided OTFS System in Doubly Dispersive Channels

Phase Optimization Approaches for RIS-Aided OTFS System in D...

引用

IEEE/CIC International Conference on Communications in China (ICCC)

作者： Xu, Ran Wang, Zhaoqi Gao, Dixiang Liu, Xiqing Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350378412

Orthogonal time frequency space (OTFS) modulation is a candidate technique for the sixth generation (6G) mobile communication system with a speed-tolerant capability. Reconfigurable intelligent surface (RIS) aided OTFS has been proposed to surmount doubly selective fading since RIS modifies the propagation characteristics by controlling the reflection features. Existing phase optimization methods for RIS-aided OTFS focus on capacity maximization, while reliability cannot be guaranteed especially in complex channels. With channel model, input-output relation and channel estimation scheme carried out, this work proposes two optimization approaches for RIS-aided OTFS, i.e., maximum variance (MV) and maximum tap magnitude (MTM). A modified continuous MTM is designed to further reduce bit error rate (BER) and computational complexity but only applicable in continuous phase shift models. Thereafter, genetic algorithm (GA) is utilized to solve the optimization problems of MV and MTM, and the closed-form solution of continuous MTM is derived. Simulation results show that the proposed approaches gain a low BER with acceptable capacity loss, present evident advantages in harsh environments, and spare time complexity to a great extent using GA.

关键词： Reconfigurable intelligent surface orthogonal time frequency space doubly dispersive channel phase optimization genetic algorithm

来源：评论

学校读者我要写书评

暂无评论

Multi-Agent Cooperative Game Based Task Computing Mechanism for UAV-Assisted 6G NTN

引用

MOBILE NETWORKS & APPLICATIONS 2023年第4期28卷 1510-1518页

作者： Shao, Sujie Su, Lili Guo, Shaoyong Yu, Peng Qiu, Xuesong Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Xitucheng Rd 10 Beijing 100876 Peoples R China

Unmanned Aerial Vehicle (UAV)-Assisted becomes promising prospects for the development of 6G non-terrestrial networks (NTN) due to their low cost, low latency, flexible coverage, and support for emergency response. UAVs utilize close-range multi-point cooperative computing to provide computational, storage, and communication resources for 6G NTN. However, insufficient resources and variable business requirements can lead to node overload or task timeout. Therefore, selecting collaborative nodes and optimizing data transmission paths between non-terrestrial nodes has become a formidable challenge. Due to the capacity constraints, power limitations, time-varying communication links, and increasing terminal demands, existing algorithms face difficulties in meeting the requirements of cooperative computing for non-terrestrial nodes. To solve the issue, this paper proposes a multi-task cooperative computing mechanism that focuses on node selection and path optimization. Furthermore, node selection and path optimization algorithm based on multi-agent proximal policy optimization (NSPO-MAPPO) is designed to optimize the latency-aware multi-task cooperative computing objective, thereby improving computational task execution efficiency and service quality. Experimental comparisons with other algorithms demonstrate the effectiveness of our method in minimizing latency, enhancing the task load balancing degree, and ensuring stable and reliable cooperative computing services for non-terrestrial nodes.

关键词： UAV-assisted 6G NTN Cooperative computing Multi-agent game Multi-task

来源：评论

学校读者我要写书评

暂无评论

Optimizing Beam Hopping in Multibeam NGSO Constellations with Multi-Agent Reinforcement Learning 7

Optimizing Beam Hopping in Multibeam NGSO Constellations wit...

引用

IEEE 7th International Workshop on Radio Frequency and Antenna Technologies (IEEE IWRF and AT)

作者： Zhang, Yifan Wang, Ying Zhang, Qiuyang Jia, Huaiqi Feng, Linqing Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350384765;9798350384772

With the rapid development of Non-geostationary orbit (NGSO) satellite constellation, user demands have significantly increased. However, due to the high mobility of satellites and the uneven distribution of users within the coverage area, there are temporal and geographical variations in service demand. This leads to a situation where some beams in the system are unable to provide the required capacity, while others offer excess capacity, resulting in inefficient resource utilization. One of the primary challenges in future constellation design is addressing these imbalanced dynamic traffic demands. This paper proposes a novel approach to managing satellite resources focusing on beam hopping by leveraging multi-agent reinforcement learning. Compared with the traditional heuristic algorithm, the long-term return can be considered to better cope with the changing traffic demand. Moreover, the proposed agent reinforcement learning method has small action space, low training complexity and high training stability when dealing with large constellations. Simulation results demonstrate that proposed algorithm can effectively meet dynamically changing demands, thereby enhancing the overall service quality of the constellation.

关键词： NGSO constellation beam hopping dynamic resource management multi-agent reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

Large Language Model Enhanced Autonomous Agents for Proactive Fault-Tolerant Edge Networks

Large Language Model Enhanced Autonomous Agents for Proactiv...

引用

IEEE Conference on Computer Communications (IEEE INFOCOM)

作者： Fang, Honglin Zhang, Di Tan, Can Yu, Peng Wang, Ying Li, Wenjing Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350384482;9798350384475

Addressing network fault tolerance tasks with various scenarios and domains is a crucial step towards achieving autonomous networks. Despite the abundance of artificial intelligence models designed for specific network scenarios and fault tasks, they may encounter challenges in delivering optimal performance across all environments. Considering the exhibited exceptional abilities of large language models (LLMs) in content generation and task planning, we advocate that LLMs can act as automated schedulers to manage existing fault tolerance models. Through the current network scenarios and task prompts, appropriate models are invoked to orchestrate optimization strategies. By leveraging pre-optimized knowledge and tools, we propose an LLM-based edge network fault-tolerant paradigm (LLM-ENFT), which autonomously manages the full cycle of perceiving, diagnosing, and recovering network faults. Experiments conducted on the edge network testbed based on Mininet and Ryu show that LLM-ENFT exhibits strong resilience in the face of network congestion and link failure.

关键词： large language model autonomous agent fault tolerance edge networks

来源：评论

学校读者我要写书评

暂无评论

Bridging Beam Split and Semantic Communications: Enhancing Transmission Efficiency

Bridging Beam Split and Semantic Communications: Enhancing T...

引用

IEEE/CIC International Conference on Communications in China (ICCC)

作者： Ma, Junyao Sun, Mengying Xu, Xiaodong Dong, Chen Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350377675;9798350377682

This paper proposes an image semantic communication scheme that utilizes entropy-based ordering under beam split transmission conditions, named BSISC. Based on the entropy model, the BSISC scheme matches semantic features of varying importance, encoded through joint source-channel coding, with the fading channels produced by beam splitting. This enhances the quality of information recovery. The BSISC scheme is expanded to a dual-user scenario, resulting in the mBSISC scheme. The mBSISC scheme can serve two users simultaneously through channel multiplexing technology based on channel differentiation, ensuring the presence of the main beam lobes without mutual interference. The mBSISC scheme effectively improves resource utilization efficiency and ensures the quality of information recovery for two users. Simulation results demonstrate that the proposed BSISC significantly outperforms methods without entropy-based ordering in terms of the peak signal-to-noise ratio of the received images. Moreover, mBSISC can serve two users simultaneously while guaranteeing recover quality.

关键词： beam split semantic communications JSCC entropy model

来源：评论

学校读者我要写书评

暂无评论

SuperLED: Supervised Contrastive Learning based Dual Path Triple Extraction Framework

SuperLED: Supervised Contrastive Learning based Dual Path Tr...

引用

International Joint Conference on Neural Networks (IJCNN)

作者： He, Luyao Zhang, Zhongbao Su, Sen Wang, Hui Beijing Univ Posts & Telecommun State Key Lab Networking & Switching Technol Beijing Peoples R China

ISBN: (纸本)9798350359329;9798350359312

Relation triple extraction is an essential task in information extraction and knowledge graph construction. Despite recent advancements, existing methods still exhibit certain limitations. Firstly, they only employ generalized pre-trained models and do not consider the specificity of relation triple extraction tasks. Moreover, existing tagging-based approaches typically decompose the task into two subtasks, initially identifying subjects and subsequently identifying objects and relations. They solely focus on extracting relational triples from subject -> object, neglecting that once the extraction of a subject fails, it fails in extracting all triples. To address these issues, we propose SuperLED, an innovative supervised contrastive learning based dual path triple extraction framework. Specifically, we design a supervised contrastive learning method that considers multiple positive samples per anchor rather than restricting it to only one positive. Furthermore, a penalty term is introduced to prevent excessive similarity between the subject and object. Our framework implements taggers in two paths, enabling triples extraction from subject -> object and object -> subject. Experimental results show that SuperLED achieves state-of-the-art results on the benchmark datasets and significantly improves the F1 score on Normal, SEO, EPO, and multiple relation extraction tasks.

关键词： Contrastive Learning

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：