检索结果-内蒙古大学图书馆

A unified framework of data augmentation using large language models for text-based cross-modal retrieval

Pattern Recognition 2025年 167卷

作者： Si, Lijia Guo, Caili Li, Zheng Yang, Yang Beijing Key Laboratory of Network System Architecture and Convergence Beijing University of Posts and Telecommunications Beijing China Beijing Laboratory of Advanced Information Networks Beijing University of Posts and Telecommunications Beijing China China Telecom Digital Intelligence Technology Co. Ltd Beijing China

Text-based cross-modal retrieval (TCMR) is an information retrieval task that aims to find semantically equivalent content from other modalities given a textual description. However, collecting well-labeled data for TCMR is very expensive and time-consuming, which highly affects the performance of TCMR. Recently, large language models (LLMs) have approached or even surpassed the human level on many natural language processing tasks, which inspires us to try to use LLMs for data augmentation of TCMR datasets. However, data augmentation using LLMs also faces significant challenges, particularly the hallucination problem of LLMs and the fusion of augmented data with the original data. In this paper, we propose LLM-DA++, a unified framework for data augmentation using large language models that is applied to text-based cross-modal retrieval models. LLM-DA++ includes the training and inference sub-frameworks, both leveraging LLM to rewrite texts, enhancing vocabulary diversity and sentence structure while retaining key semantics. In order to alleviate the hallucinations problem of LLMs, a Text Faithfulness Filter is introduced in LLM-DA++ to filter out unfaithful rewritten texts. In addition, a Balanced Sampling Strategy is introduced in LLM-DA++ to balance the contributions of original and augmented data. Comprehensive experiments on three cross-modal retrieval tasks, text-based audio retrieval, text-based person retrieval, and text-based motion retrieval, show that LLM-DA++ can significantly improve the retrieval performance of existing TCMR tasks, validating the effectiveness of LLM-DA++. The code is available at https://***/slj5/LLM-DA-plus-plus. © 2025

关键词： Content based retrieval

来源：评论

学校读者我要写书评

暂无评论

Coded Caching for Smart Grid Enabled HetNets With Resource Allocation and Energy Cooperation

引用

IEEE TRANSACTIONS ON VEHICULAR TECHNOLOGY 2020年第10期69卷 12058-12071页

作者： Yin, Fangfang Zeng, Minyin Zhang, Zhilong Liu, Danpu Commun Univ China State Key Lab Media Convergence & Commun Beijing 100024 Peoples R China Beijing Univ Posts & Telecommun Beijing Lab Adv Informat Network Beijing Key Lab Network Syst Architecture & Conve Beijing 100876 Peoples R China

With the development of millions more small-cell base stations (SBSs) and billions of connected devices in the future wireless networks, the need for energy-efficient system design and operation will be even more compelling. This paper investigates the content caching, resource allocation and energy cooperation in coded cache enabled heterogeneous networks (HetNets), where SBSs with cache ability are powered by both conventional grids and renewable energy sources, and energy can be shared among SBSs via the smart grid. We formulate the joint optimization problem, aims at minimizing the conventional grid energy consumption while satisfying quality of service (QoS) requirements of users. To deal with the mixed integer non-linear optimization problem, the original problem is decomposed into three lower level problems, i.e., content caching, content delivery and energy cooperation. We first propose an alternative iterative algorithm to optimize content caching and content delivery by jointly applying convex optimization techniques and matching theory. Then, by discussing two representative energy cooperative scenarios of smart grid supported networks, we propose efficient energy cooperation algorithms for the two scenarios, respectively. Furthermore, the joint content caching, content delivery and energy cooperation optimization algorithms are developed to minimize the conventional grid energy consumption in the considered HetNets. Simulation results demonstrate that the joint algorithms can significantly reduce the conventional grid energy consumption compared with other benchmarks.

关键词： Coded caching energy cooperation smart grid matching bandwidth allocation user association

来源：评论

学校读者我要写书评

暂无评论

An SDN/NFV Based Framework for Management and Deployment of Service Based 5G Core network

引用

China Communications 2018年第10期15卷 86-98页

作者： Lu Ma Xiangming Wen Luhan Wang Zhaoming Lu Raymond Knopp Beijing Key Laboratory of Network System Architecture and Convergence Beijing University of Posts and Telecommunications Beijing 100876 China Beijing Advanced Innovation Center for Future Internet Technology Beijing 100124 China Communication System Department EURECOM Biot 06410 France

The traffic explosion and the rising of diverse requirements lead to many challenges for traditional mobile network architecture on flexibility, scalability, and deployability. To meet new requirements in the 5 G era, service based architecture is introduced into mobile networks. The monolithic network elements(e.g., MME, PGW, etc.) are split into smaller network functions to provide customized services. However, the management and deployment of network functions in service based 5 G core network are still big challenges. In this paper, we propose a novel management architecture for 5 G service based core network based on NFV and SDN. Combined with SDN, NFV and edge computing, the proposed framework can provide distributed and on-demand deployment of network functions, service guaranteed network slicing, flexible orchestration of network functions and optimal workload allocation. Simulations are conducted to show that the proposed framework and algorithm are effective in terms of reducing network operating cost.

关键词： service based architecture 5G core network SDN NFV workload allocation

来源：评论

学校读者我要写书评

暂无评论

Mobility-Based Proactive Cache and Transmission Strategy in Millimeter Wave Heterogeneous networks

引用

IEEE systemS JOURNAL 2023年第1期17卷 325-336页

作者： Yin, Fangfang Zhang, Zhilong Liu, Danpu Li, Pengru Zeng, Minyin Commun Univ China State Key Lab Media Convergence & Commun Beijing 100024 Peoples R China Beijing Univ Posts & Telecommun Beijing Lab Adv Informat Network Beijing Key Lab Network Syst Architecture & Conve Beijing 100876 Peoples R China

Millimeter wave (mmWave) frequencies with rich bandwidth can greatly alleviate the surging traffic demand of wireless communication from mobile video streaming. However, the harsh propagation environment at mmWave bands and user mobility make it hard to provide a reliable service. Multibeam transmissions can be taken as an essential approach for maintaining a robust mmWave link. Meanwhile, the introduction of coded caching can further reduce the end-to-end latency and backhaul traffic. Considering the interaction between caching and transmission in mobility scenario, we develop a joint design for mmWave heterogeneous networks (HetNets), by combining multibeam transmission and maximum distance separable (MDS)-coded caching. First, a joint optimization problem of cache placement and beam pair selection is formulated, in order to maximize the backhaul traffic offload in mmWave HetNets. To solve the mixed-timescale mixed-integer nonlinear programming (MINLP) problem, we then decompose the joint problem into the long-term mobility-based proactive content caching (CC) and short-term content delivery (CD) subproblems. Subsequently, two many-to-many matching-based algorithms are proposed for tackling the CD subproblem and the convex technique is applied to deal with the CC subproblem. Simulation results demonstrate the efficacy of the proposed algorithms over benchmark schemes.

关键词： Millimeter wave communication Videos Robustness Interference Quality of service Optimization laboratories Beam pair selection coded caching matching millimeter wave (mmWave) mobility

来源：评论

学校读者我要写书评

暂无评论

Hybrid field physical layer network coding in multi-user relay network

Hybrid field physical layer network coding in multi-user rel...

引用

National Doctoral Academic Forum on Information and Communications Technology 2013

作者： Guo, Yijun Hao, Jianjun Yue, Guangxin Beijing Key Laboratory of Network System Architecture and Convergence Beijing University of Posts and Telecommunications China

ISBN: (纸本)9781849198196

Physical layer network coding (PNC) exploits the broadcast nature of wireless channel to obtain transmission efficiency gains. This paper proposes a hybrid field physical layer network coding (HFNC) approach and applys HFNC to the multi-user relay networks. The proposed method enables a reliable transmission with variable throughput under different signal-to-noise ratios (SNR). With given constellation rotations at sources, the points of the equivalent superpositioned (ESP) constellation cluster. The HFNC-based decoding can be divided into two layers. Point-level (PL) decoding over the complex-field achieves a higher throughput as complex field network coding (CFNC), while cloud-level decoding defined in the finite-field is able to resist bad channel with a lower throughput. Excellent performances of symbol error probability (SEP) and transmission efficiency are proved by both theoretical analysis and simulation results.

关键词： Throughput

来源：评论

学校读者我要写书评

暂无评论

Deep learning for channel decoding under correlated noise

Deep learning for channel decoding under correlated noise

引用

2019 International Conference on Advances in Computer Technology, Information Science and Communications, CTISC 2019

作者： Zhang, Xiangxiang Luo, Tao Beijing Key Laboratory of Network System Architecture and Convergence Beijing University of Posts and Telecommunications China

ISBN: (纸本)9789897583575

Traditional Viterbi decoding algorithm degrades performance under correlated noise. Inspired by the powerful learning from data of deep neural networks, we apply denoising convolutional neural network (DnCNN) to channel decoding under correlated noise. In this paper, we propose a DnCNN hard decision Viterbi decoder (short for DHDV decoder) architecture to enhance the performance of convolutional codes under correlated noise. The architecture applies DnCNN to denoise and improves the decoding SNR (signal to noise ratio) through the correlation of noise correlation. Then, the Viterbi Algorithm decoder decodes from the denoised data. DHDV decoder obtains greater BER performance improvement under stronger noise correlation, and it is robust under different convolutional codes and correlated noise models. Comparing the complexity of DnCNN network and matrix multiplication whitening method, DnCNN network complexity is lower when the code length is longer. © 2019 by SCITEPRESS - Science and Technology Publications, Lda.

关键词： Viterbi algorithm

来源：评论

学校读者我要写书评

暂无评论

A novel ser and PAPR reduction scheme for complex field network coding 10

A novel ser and PAPR reduction scheme for complex field netw...

引用

10th International Conference on Wireless Communications, networking and Mobile Computing, WiCOM 2014

作者： Li, Pan Hao, Jianjun Guo, Yijun Beijing Key Laboratory of Network System Architecture and Convergence Beijing University of Posts and Telecommunications Beijing China

ISBN: (纸本)9781849198455

For multi-source relay-based cooperative communications, it is widely believed that Complex Field network Coding (CFNC) achieves better throughput performance than Physical-layer network Coding (PNC). However, traditional CFNC schemes suffer a considerable loss on the performance of symbol to error ratio (SER) and peak-toaverage power ratio (PAPR) as the number of source nodes increases. In this paper, to tackle this problem in some extent, a novel CFNC scheme, which comprises of two SER and PAPR reduction algorithms, is proposed and analyzed. Particularly, by optimizing the complex coefficient vector at the source nodes at first, the performance of SER is improved;while by shrinking the amplitude of some signals selectively, the PAPR at the destination nodes is reduced. Simulations are provided to validate our analytical results. From the simulation results, it is observed that the proposed CFNC scheme significantly improves the performance of SER and PAPR.

关键词： network coding

来源：评论

学校读者我要写书评

暂无评论

Machine Learning Inspired Binocular Vision Indoor Positioning system Based on QR Code Beacon 2020

Machine Learning Inspired Binocular Vision Indoor Positionin...

引用

4th International Conference on Innovation in Artificial Intelligence, ICIAI 2020

作者： Deng, Xinyi He, Dazhong Zhou, Wenli Yang, Yang Zhang, Yue Beijing Key Laboratory of Network System Architecture and Convergence Beijing University of Posts and Telecommunications Beijing China

ISBN: (纸本)9781450376587

With the development of the Internet of Things, indoor positioning technology has a wider application prospect. Binocular vision positioning technology can effectively meet the needs of high positioning accuracy and low hardware costs. Based on the traditional binocular vision ranging process, we aims to improve the ranging and positioning accuracy of the binocular system by using machine learning (ML) methods and Quick Response (QR) code beacons. First, this article will use the binarization technology to locate the QR code area. Second, using the fully convolutional network (FCN) structure, we design a convolutional neural network (CNN) model and train it on the KITTI dataset, and convolve the QR code region to get the disparity map of the region. Third, in order to accurately locate the coordinate position of the beacon center, our proposed system uses a density-based clustering algorithm to cluster 3D point clouds, and uses binary linear regression to fit the plane equations of the QR code beacon. Finally, the positioning experiments are carried out in practical applications. The experimental results show that compared with the traditional binocular vision algorithm, the system has higher positioning accuracy, and has performance close to the theoretical accuracy of binocular vision. © 2020 ACM.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Information extraction of forum based on regular expression

Information extraction of forum based on regular expression

引用

2013 5th International Conference on Intelligent Human-Machine systems and Cybernetics, IHMSC 2013

作者： He, Gang Zhang, Yingwei Wu, Xiaochun Beijing Key Laboratory of Network System Architecture and Convergence Beijing University of Posts and Telecommunications Beijing China

ISBN: (纸本)9780769550114

This paper introduces the popular universal forum systems in domestic mainstream forum and analyzes the unique characteristics of these forum systems. Based on these unique characteristics, we propose the concept of system fingerprint which can used to detect the different systems of forum exactly and extract the users' information efficiently. It contributes to the development of network information auditing. Experimental results show that the approach can achieve high extraction accuracy. It has important application value and practical significance. © 2013 IEEE.

关键词： Pattern matching

来源：评论

学校读者我要写书评

暂无评论

A bio-inspired approach for network selection in dense heterogeneous networks

A bio-inspired approach for network selection in dense heter...

引用

IEEE Asia Pacific Conference on Wireless and Mobile, APWiMob 2015

作者： Chen, Yawen Wen, Xiangming Lu, Zhaoming Shao, Hua Cheng, Jie Xi, Zeguo Beijing Key Laboratory of Network System Architecture and Convergence Beijing University of Posts and Telecommunications Beijing China

ISBN: (纸本)9781479982905

The emerging dense heterogeneous networks comprising different types of small cells enable consumers to employ multiple radio access technologies to communicate. In such a dymica network environment, development of robust and self-adaptive network selection mechanisms is required to satisfy the growing demand on ubiquitous network access. Besides, the energy consumed at wireless interface accounts for the great proportion of the total energy consumption of user devices and the battery lifetime has become a important factor affecting quality of experience (QoE). Hence in this paper, we propose an adaptive network selection approach where each user autonomously choose desired wireless network interface that offers the best energy-quality tradeoff. For this purpose, we adopt an attractor selection model, which is developed from autonomous and adaptive behavior of biological systems. The high adaptability to fluctuating network environment and the simple control rules allow users to dynamically select the best network that provide satisfactory QoE while achieve a acceptable device battery lifetime. © 2015 IEEE.

关键词： Quality of service

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：