检索结果-南通市图书馆

Compositional metric learning for multi-label classification

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

Frontiers of Computer Science 2021年第5期15卷 1-12页

作者： Yan-Ping SUN Min-Ling ZHANG School of Computer Science and Engineering Southeast UniversityNanjing 210096China Key Laboratory of Computer Network and Information Integration(Southeast University) Ministry of EducationChina collaborative innovation center for wireless communications technology Nanjing 211100China

Multi-label classification aims to assign a set of proper labels for each instance,where distance metric learning can help improve the generalization ability of instance-based multi-label classification *** multi-label metric learning techniques work by utilizing pairwise constraints to enforce that examples with similar label assignments should have close distance in the embedded feature *** this paper,a novel distance metric learning approach for multi-label classification is proposed by modeling structural interactions between instance space and label *** one hand,compositional distance metric is employed which adopts the representation of a weighted sum of rank-1 PSD matrices based on com-ponent *** the other hand,compositional weights are optimized by exploiting triplet similarity constraints derived from both instance and label *** to the compositional nature of employed distance metric,the resulting problem admits quadratic programming formulation with linear optimization complexity *** number of training *** also derive the generalization bound for the proposed approach based on algorithmic robustness analysis of the compositional *** experiments on sixteen benchmark data sets clearly validate the usefulness of compositional metric in yielding effective distance metric for multi-label classification.

关键词： machine learning multi-label learning metric learning compositional metric positive semidefinite matrix decomposition

Binary relevance for multi-label learning： an overview

在线全文

学校读者我要写书评

暂无评论

Frontiers of Computer Science 2018年第2期12卷 191-202页

作者： Min-Ling ZHANG Yu-Kun LI Xu-Ying LIU Xin GENG School of Computer Science and Engineering Southeast University Nanjing 210096 China Key Laboratory of Computer Network and Information Integration (Southeast University) Ministry of Education China collaborative innovation center for wireless communications technology Nanjing 211100 China

Multi-label learning deals with problems where each example is represented by a single instance while being associated with multiple class labels simultaneously. Binary relevance is arguably the most intuitive solution for learning from multi-label examples. It works by decomposing the multi-label learning task into a number of independent binary learning tasks （one per class label）. In view of its potential weakness in ignoring correlations between labels, many correlation-enabling extensions to binary relevance have been proposed in the past decade. In this paper, we aim to review the state of the art of binary relevance from three perspectives. First, basic settings for multi-label learning and binary relevance solutions are briefly summarized. Second, representative strategies to provide binary relevance with label correlation exploitation abilities are discussed. Third, some of our recent studies on binary relevance aimed at issues other than label correlation exploitation are introduced. As a conclusion, we provide suggestions on future research directions.

关键词： machine learning multi-label learning binary relevance label correlation class-imbalance relative labeling-importance

Low Complexity SOVA for Turbo Codes

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

China communications 2017年第8期14卷 33-40页

作者： Ruilin Pei Zulin Wang Qin Huang Jiaqing Wang School of Electronic and Information Engineering Beihang UniversityBeijing100191China collaborative innovation center of Geospatial technology Wuhan430079China State Key Laboratory of wireless Mobile communications(CATT) Beijing100191China

Recently, trimming Soft-output Viterbi algorithm(T-SOVA) has been proposed to reduce the complexity of SOVA for Turbo codes. In its fi rst stage, a dynamic algorithm, lazy Viterbi algorithm, is used to indicate the minimal metric differences which brings obstacle on hardware implementation. This paper proposes a Viterbi algorithm(VA) based T-SOVA to facilitate hardware implementation. In the first stage of our scheme, a modified VA with regular structure is used to fi nd the maximum likelihood(ML) path and calculate the metric differences. Further, local sorting is introduced to trim the metric differences, which reduces the complexity of trimming operation. Simulation results and complexity analysis show that VA based T-SOVA performs as well as lazy VA based T-SOVA and is easier to be applied to hardware implementation.

关键词： VA SOVA local sorting hardware implementation

Transfer synthetic over-sampling for class-imbalance learning with limited minority class data

在线全文

学校读者我要写书评

暂无评论

Frontiers of Computer Science 2019年第5期13卷 996-1009页

作者： Xu-Ying LIU Sheng-Tao WANG Min-Ling ZHANG School of Computer Science and Engineering Southeast University Nanjing 210096 China Key Laboratory of Computer Network and Information Integration (Southeast University) Ministry of Education Nanjing 210096 China collaborative innovation center for wireless communications technology Nanjing 210096 China

The problem of limited minority class data is encountered in manyclass imbalanced applications, but has received little attention. Synthetic over-sampling, as popular class-imbalance learning methods, could introduce much noise when minority class has limited data since the synthetic samples are not i.i.d. samples of minority class. Mos t sophisticated synthetic sampling methods tackle this problem by denoising or generating samples more consistent with ground-truth data distribution. But their assumptions about true noise or ground-truth data distribution may not hold. To adapt synthetic sampling to the problem of limited minority class data, the proposed Traso framework treats synthetic minority class samples as an additional data source, and exploits transfer learning to transfer knowledge from them to minority class. As an implementation, TrasoBoost method firstly generates synthetic samples to balance class sizes. Then in each boosting iteration, the weights of synthetic samples and original data decrease and increase respectively when being misclassified, and remain unchanged otherwise. The misclassified synthetic samples are potential noise, and thus have smaller influence in the following iterations. Besides, the weights of minority class instances have greater change than those of majority class instances to be more influential. And only original data are used to estimate error rate to be immune from noise. Finally, since the synthetic samples are highly related to minority class, all of the weak learners are aggregated for prediction. Experimental results show TrasoBoost outperforms many popular class-imbalance learning methods.

关键词： machine learning data mining class imbalance over sampling boosting transfer learning

A novel fractional uplink power control framework for self-organizing networks

维普期刊数据库评论

在线全文

维普期刊数据库

学校读者我要写书评

暂无评论

Digital communications and Networks 2023年第6期9卷 1434-1440页

作者： Zezhou Luo Hongcheng Zhuang wireless technology innovation center Guangdong Communications and Networks InstituteShenzhen510663China School of Electronics and Communication Engineering SUN YAT-SEN UniversityShenzhen518107China

Internet of things and network densification bring significant challenges to uplink *** depending on optimization algorithm enhancements is not enough for uplink *** control intercell interference,Fractional Uplink Power Control(FUPC)should be optimized from network-wide perspective,which has to find a better traffic distribution ***,traffic distribution is geographic-based,and ineffective due to tricky locating *** paper proposes a novel uplink power management framework for Self-Organizing Networks(SON),which firstly builds up pathloss-based traffic distribution model and then makes the decision of FUPC based on the ***-based Traffic Distribution(PLTD)aggregates traffic based on the propagation condition of traffic that is defined as the pathloss between the position generating the traffic and surrounding *** show that the improvement in optimization efficiency of FUPC with PLTD can be up to 40%compared to conventional GeoGraphic-based Traffic Distribution(GGTD).

关键词： 5G and beyond Self-organizing networks Uplink power control Optimization efficiency Traffic distribution

Space-Time Correlation Based Fast Regional Spectrum Sensing in Cognitive Radio

在线全文

学校读者我要写书评

暂无评论

China communications 2017年第5期14卷 78-90页

作者： Sai Huang Yuanyuan Yao Zhiyong Feng Ping Zhang Yifan Zhang Key Laboratory of Universal wireless communications Ministry of Education wireless technology innovation Institute(WTI) Beijing University of Posts and TelecommunicationsBeijing 100876China

In this paper,a space-time correlation based fast regional spectrum sensing(RSS)scheme is proposed to reduce the time and energy consumption of traditional spatial spectrum sensing. The target region is divided into small meshes,and all meshes are clustered into highly related groups using the spatial correlation among them. In each group,some representative meshes are selected as detecting meshes(DMs)using a multi-center mesh(MCM)clustering algorithm,while other meshes(EMs)are estimated according to their correlations with DMs and the Markov modeled dependence on history by MAP principle. Thus,detecting fewer meshes saves the sensing consumption. Since two independent estimation processes may provide contradictory results,minimum entropy principle is adopted to merge the results. Tested with data acquired by radio environment mapping measurement conducted in the downtown Beijing,our scheme is capable to reduce the consumption of traditional sensing method with acceptable sensing performance.

关键词： Cognitive radio Channel occupancy estimation Dynamic spectrum access Radio environment mapping Space-time correlation Spectrum sensing

Low-Profile Dual-Polarized Planar Antenna with Compact Structure for Base Stations

在线全文

学校读者我要写书评

暂无评论

Chinese Journal of Electronics 2017年第5期26卷 1092-1095页

作者： LIU Ying WANG Jiu GONG Shuxi Science and technology on Antenna and Microwave Laboratory Collaborative Innovation Center of Information Sensing and UnderstandingXidian University

A low-profile dual-polarized planar antenna with compact structure for 2.4GHz wireless local area network(WLAN) and 2.5GHz Long term evaluation(LTE2500) base stations is proposed.A combination of a planar radiator and compact Artificial magnetic conductor(AMC) reflector is utilized to achieve a full planar structure without adding extra matching circuit,an extremely low-profile of 0.127 wavelength,as well as a high front to back ratio(F/B) of 25 *** antenna is fabricated to test the performance such as S-parameters,radiation patterns and peak *** results show the suitability of the proposed antenna for 2.4GHz WLAN and LTE2500 applications.

关键词： Low profile Planar Compact structure Base station antenna

Open Access Strategy in Cloud Computing-Based Heterogenous Networks Constrained by wireless Fronthaul

在线全文

同方期刊数据库

学校读者我要写书评

暂无评论

China communications 2017年第2期14卷 143-153页

作者： Zhongyuan Zhao Yong Li Weiliang Xie Min Wang Wenbo Wang Key Laboratory of Universal wireless communications (Ministry of Education) Beijing University of Posts and TelecommunicationsBeijing 100876 China China Telecom technology innovation center Beijing 102209 China

Cloud *** heterogeneous networks(*** Nets) have been considered as promising network architecture of future wireless communication systems. However, its performance is constrained by the capability of fronthaul. To study the performance limit of *** Nets, the coverage performance of the open access strategy in *** Nets is analyzed in this paper. In particular, the *** condition of fronthaul in *** Nets is characterized by establishing a stochastic *** model, and a tight lower bound on coverage probability is derived for downlink transmissions. Based on the derived results, the constraint of the fronthaul is given to ensure that open access strategy can achieve the best coverage performance. Finally, the simulation results are provided to verify the theoretical results and show the impact of fronthaul constrains.

关键词： C-Het Nets open access strategy coverage probability stochastic geometry

Multicell power allocation method based on game theory for inter-cell interference coordination

在线全文

学校读者我要写书评

暂无评论

Science in China(Series F) 2009年第12期52卷 2378-2384页

作者： ZHANG Hui XU XiaoDong LI JingYa TAO XiaoFeng ZHANG Ping SVENSSON Tommy BOTELLA Carmen wireless technology innovation Institute Beijing University of Posts and Telecommunications Beijing 100876 China Key Laboratory of Universal wireless communications Ministry of Education Beijing 100876 China Chalmers University of technology Sweden

As a new technology, coordinated multipoint （CoMP） transmission is included in LTE-Advanced study item. Moreover, the network architecture in LTE-Advanced system is modified to take into account coordinated transmission. Under this background, a novel power allocation game model is established to mitigate inter-cell interference with cellular coordination. In the light of cellular cooperation relationship and centralized control in eNodeB, the power allocation in each served antenna unit aims to make signal to interference plus noise ratio （SINR） balanced among inter-cells. Through the proposed power allocation game algorithm, the users＇ SINR can reach the Nash equilibrium, making it feasible to reduce the co-frequency interference by decreasing the transmitted power. Numerical results show that the proposed power allocation algorithm improves the throughput both in cell-center and cell-edge. Moreover, the blocking rate in cell-edge is reduced too.

关键词： LTE-Advanced inter-cell interference coordination CoMP game theory power allocation