检索结果-内蒙古大学图书馆

SGCL: Semi-supervised Graph Contrastive Learning with confidence propagation algorithm for node classification

KNOWLEDGE-BASED SYSTEMS 2024年 301卷

作者： Jiang, Wenhao Bai, Yuebin Beihang Univ Sch Comp Sci & Engn Beijing 100191 Peoples R China

Semi-Supervised Graph Learning (SSGL) aims to predict massive unknown labels based on a subset of known labels within a graph. Recently, graph neural network, one of the most popular SSGL approaches, has garnered considerable research interest and achieved remarkable progress. However, many of these methods perform unsatisfactorily with limited labeled data. Graph contrastive learning (GCL), which utilizes unlabeled data to generate supervision, partially addresses this issue but does not fully exploit label information. To address this challenge, we propose SSGL algorithm, Semi-supervised Graph Contrastive Learning with Confidence propagation algorithm (SGCL). SGCL comprises two stages of contrastive learning. In the first stage, we employ unsupervised contrastive learning to initialize the model with graph augmentation. In the second stage, in order to fully leverage known labels and graph structure, we incorporate supervised contrastive learning which utilizes supervision signals obtained from confidence propagation algorithm. By combining supervised contrastive learning and unsupervised contrastive learning, the embedding quality and the classification accuracy can be further enhanced. At last, comprehensive experiments demonstrate that SGCL outperforms the best baseline method by an average of 2.23% across six datasets, highlighting the effectiveness of our approach.

关键词： Graph Contrastive Learning label propagation algorithm Node classification Pseudo label

来源：评论

学校读者我要写书评

暂无评论

AnnoFin-A hybrid algorithm to annotate financial text

引用

EXPERT SYSTEMS WITH APPLICATIONS 2017年 88卷 270-275页

作者： Das, Ananda Swarup Mehta, Sameep Subramaniam, L. V. IBM India Res Labs New Delhi India

In this work, we study the problem of annotating a large volume of Financial text by learning from a small set of human-annotated training data. The training data is prepared by randomly selecting some text sentences from the large corpus of financial text. Conventionally, bootstrapping algorithm is used to annotate large volume of unlabeled data by learning from a small set of annotated data. However, the small set of annotated data have to be carefully chosen as seed data. Thus, our approach is a digress from the conventional approach of bootstrapping as we let the users randomly select the seed data. We show that our proposed algorithm has an accuracy of 73.56% in classifying the financial texts into the different categories ("Accounting", "Cost", "Employee", "Financing", "Sales", "Investments", "Operations", "Profit", "Regulations" and "Irrelevant") even when the training data is just 30% of the total data set. Additionally, the accuracy improves by an approximate average of 2% for an increase of the training data by 10% and the accuracy of our system is 77.91% when the training data is about 50% of the total data set. As a dictionary of hand chosen keywords prepared by domain experts are often used for financial text extraction, we assumed the existence of almost linearly separable hyperplanes between the different classes and therefore, we have used Linear Support Vector Machine along with a modified version of label propagation algorithm which exploits the notion of neighborhood (in Euclidean space) for classification. We believe that our proposed techniques will be of help to Early Warning Systems used in banks where large volumes of unstructured texts need to be processed for better insights about a company. (C) 2017 Elsevier Ltd. All rights reserved.

关键词： Text classification Financial sentences label propagation algorithm SVM Clustering

来源：评论

学校读者我要写书评

暂无评论

An Effective Identification Technology for online News Con sent Spammers in Internet Media

引用

IEEE ACCESS 2019年 7卷 37792-37806页

作者： Si, Huayou Sun, Wen Zhang, Jilin Wan, Jian Xiong, Neal N. Zhou, Li Ren, Yongjian Hangzhou Dianzi Univ Sch Comp Sci & Technol Hangzhou 310018 Zhejiang Peoples R China Hangzhou Dianzi Univ Minist Educ Key Lab Complex Syst Modeling & Simulat Hangzhou 310018 Zhejiang Peoples R China Zhejiang Univ Sci & Technol Sch Informat & Elect Engn Hangzhou 310023 Zhejiang Peoples R China Tianjin Univ Coll Intelligence & Comp Tianjin 300072 Peoples R China

With the development of mobile Internet, it is changing the way we communicate with others. Internet media have gradually become the main mobile crowdsourcing applications for information dissemination and user communication, including online news and social networks. However, the potential business opportunities have stimulated the emergence of a large number of spammers, who release false speech, advertisements, pornographic contents, and phishing websites on the media to gain commercial benefits, which seriously affects the experience of normal users. Therefore, in order to reduce the harm of false information, the research on the identification technology of spammers has been carried out extensively. However, the traditional technologies of identifying spammers involve high data costs and poor effects, and most of them are concentrated in the field of social networks, while less research is carried out in the field of online news. In this paper, we propose an effective technology of identifying online news comment spammers based on the label propagation algorithm (LPA), making full use of the user comment behaviors and contents. First of all, we collect a large amount of news and comments from NetEase News and label some users in the data as spammers or normal users manually to construct a labeled dataset. Then, a set of behavioral and semantic features are extracted and quantified from the user comment behaviors and comment contents by statistical analysis. Next, we propose the identification technology based on the LPA. Finally, the set of feature values is input into the proposed technology in different combinations, and experiments and evaluations are carried out to determine the most effective combination of features and improve the technology. The results show that the technology proposed in this paper involves a lower data cost but a better identification effect than some traditional technologies based on the supervised classifier.

关键词： Spammer identification Internet media online news comment label propagation algorithm mobile crowdsourcing applications

来源：评论

学校读者我要写书评

暂无评论

Multitask Protein Function Prediction through Task Dissimilarity

引用

IEEE-ACM TRANSACTIONS ON COMPUTATIONAL BIOLOGY AND BIOINFORMATICS 2019年第5期16卷 1550-1560页

作者： Frasca, Marco Bianchi, Nicolo Cesa Univ Milan Dipartimento Informat Via Comelico 39-41 I-20135 Milan Italy

Automated protein function prediction is a challenging problem with distinctive features, such as the hierarchical organization of protein functions and the scarcity of annotated proteins for most biological functions. We propose a multitask learning algorithm addressing both issues. Unlike standard multitask algorithms, which use task (protein functions) similarity information as a bias to speed up learning, we show that dissimilarity information enforces separation of rare class labels from frequent class labels, and for this reason is better suited for solving unbalanced protein function prediction problems. We support our claim by showing that a multitask extension of the label propagation algorithm empirically works best when the task relatedness information is represented using a dissimilarity matrix as opposed to a similarity matrix. Moreover, the experimental comparison carried out on three model organism shows that our method has a more stable performance in both "protein-centric" and "function-centric" evaluation settings.

关键词： Proteins Protein engineering Prediction algorithms Symmetric matrices labeling Standards Ontologies Multitask learning protein function prediction label propagation algorithm gene ontology task dissimilarity

来源：评论

学校读者我要写书评

暂无评论

Broccoli: Combining Phylogenetic and Network Analyses for Orthology Assignment

引用

MOLECULAR BIOLOGY AND EVOLUTION 2020年第11期37卷 3389-3396页

作者： Derelle, Romain Philippe, Herve Colbourne, John K. Univ Birmingham Sch Biosci Birmingham W Midlands England UMR CNRS 5321 Stn Ecol Theor & Expt Moulis France Univ Montreal Ctr Robert Cedergren Dept Biochim Montreal PQ Canada

Orthology assignment is a key step of comparative genomic studies, for which many bioinformatic tools have been developed. However, all gene clustering pipelines are based on the analysis of protein distances, which are subject to many artifacts. In this article, we introduce Broccoli, a user-friendly pipeline designed to infer, with high precision, orthologous groups, and pairs of proteins using a phylogeny-based approach. Briefly, Broccoli performs ultrafast phy-logenetic analyses on most proteins and builds a network of orthologous relationships. Orthologous groups are then identified from the network using a parameter-free machine learning algorithm. Broccoli is also able to detect chimeric proteins resulting from gene-fusion events and to assign these proteins to the corresponding orthologous groups. Tested on two benchmark data sets, Broccoli outperforms current orthology pipelines. In addition, Broccoli is scalable, with runtimes similar to those of recent distance-based pipelines. Given its high level of performance and efficiency, this new pipeline represents a suitable choice for comparative genomic studies. Broccoli is freely available at https://***/rderelle/Broccoli.

关键词： orthology orthologous groups label propagation algorithm LPA gene fusions

来源：评论

学校读者我要写书评

暂无评论

RESEARCH ON THE algorithm OF COMMUNITY DISCOVERY BASED ON THE STANDARD DEVIATION OF EDGE - BETWEENNESS 13

RESEARCH ON THE ALGORITHM OF COMMUNITY DISCOVERY BASED ON TH...

引用

13th IEEE International Computer Conference on Wavelet Active Media Technology and Information Processing (ICCWAMTIP)

作者： Tang, Yanling Li, Jianping Zhao, Gaofeng Wang, Yangyang Univ Elect Sci & Technol China Sch Comp Sci & Engn Chengdu 611731 Sichuan Peoples R China

ISBN: (纸本)9781509061266

In this paper, we mainly do the research about the non-overlapping community discovery. We use the GN algorithm and LP algorithm to test the randomly generated network, and compare the two algorithms in two layers, which are the time complexity level, and the static network's synchronization capability level. The simulation results show that the two algorithms are applicable to different scenes.

关键词： Community discovery Girvan-Newman algorithm label propagation algorithm Standard deviation of edge-betweenness

来源：评论

学校读者我要写书评

暂无评论

A text-based mining approach for real estate policy impact monitoring and analysis

A text-based mining approach for real estate policy impact m...

引用

9th IEEE International Conference on Big Data (IEEE BigData)

作者： Cao, Lei Xu, Peng Shang, Wei Chinese Acad Sci Acad Math & Syst Sci Beijing Peoples R China Vivo Mobile Commun Shenzhen Co Ltd Shenzhen Peoples R China

ISBN: (纸本)9781665439022

Text mining has been widely used in economic analysis. Besides extracting individual opinion from a single text, it is more important to accurately aggregate group opinions from mass text files. This research proposed a text mining framework for constructing policy impact indicators to monitoring the public opinions on real estate market policies. A real estate domain dictionary has been developed based on the released policy announcements in Chinese real estate government websites. And corresponding public opinion index are constructed to evaluate the impact of real estate policies on the target market. In addition, we have improved the handling of emotional words in the public opinion dictionary. Word2Vec and So-PMI models are used to identify the polarity of trend words, and the score of each sentiment words is calculated based on its co-occurrences with the trends words using a modified label propagation algorithm. Beijing real estate market is taken as the test bed. Empirical study show that the policy public opinion index explains a significant part of Beijing housing price changes. In-sample and out-of-sample forecasts of Beijing housing prices indicate that the policy public opinion index is capable of increase the prediction accuracy.

关键词： text mining policy public opinion index real estate market macroeconomic policy integrated learning label propagation algorithm

来源：评论

学校读者我要写书评

暂无评论

Graph Neural Network with Neighborhood Reconnection 1

引用

16th International Conference on Knowledge Science, Engineering and Management (KSEM)

作者： Guo, Mengying Sun, Zhenyu Wang, Yuyi Liu, Xingwu Chinese Acad Sci Inst Comp Technol Beijing Peoples R China Univ Chinese Acad Sci Beijing Peoples R China CRRC Zhuzhou Inst Zhuzhou Peoples R China Dalian Univ Technol Sch Math Sci Dalian Peoples R China

ISBN: (数字)9783031402838

ISBN: (纸本)9783031402821;9783031402838

Graph Neural Network (GNN) models have become increasingly popular for network analysis, especially in node classification tasks. However, the effectiveness of GNNs is compromised by two limitations. First, they implicitly assume that networks are homophilous, leading to decreased performance on heterophilous or random networks commonly found in the real world. Second, they tend to ignore the known node labels, inferring node labels merely from the node features and network structure. This is mainly rooted in the non-uniformity of node degrees, which makes it hard to directly aggregate label information. Hence, we propose a novel framework NRGNN, short for Graph Neural Network with Neighborhood Reconnection. NRGNN adjusts the network structure to increase homophily and uniformize node degrees. Then it applies message-passing-based or PageRank-based GNNs to the reconstructed networks, addressing the two limitations of GNNs. We evaluate NRGNN against 14 state-of-the-art baselines for node classification tasks. The empirical results demonstrate that NRGNN outperforms almost all the baselines, regardless of whether the datasets are homophilous, heterophilous, or random. The techniques might be adapted to more network analysis tasks.

关键词： Graph Neural Network Heterophilous Graph Network Reconnection label propagation algorithm Node Classification

来源：评论

学校读者我要写书评

暂无评论

Community Detection and Analysis in Air Pollution Complex Network Based on LPA 30

Community Detection and Analysis in Air Pollution Complex Ne...

引用

30th International Conference on Computer Communications and Networks (ICCCN)

作者： Lian, Jie Ding, Xin Pan, Jianguo Shanghai Normal Univ Shanghai Engn Res Ctr Intelligent Educ & Bigdata Shanghai 200234 Peoples R China

ISBN: (纸本)9780738113302

Air pollution has caused a bad effect on the natural environment and public health. With the rapid establishment of air monitoring stations and meteorological stations, the volume of air pollution data is increasing dramatically. The traditional methods to model the air pollution transmission process have shown a limitation to process a large amount of non-linear data. Machine learning methods have recently been proven to be extremely efficient and accurate to process data in complex structures. In this paper, a novel method based on a complex network is proposed to discover the air monitoring station communities and analyze the air pollutants diffusion. The method comprises several aspects, including a dimension reduction process, a complex network generation process, and a community detection process. The proposed method was verified based on a real-world air pollution dataset, and the results demonstrate that the air pollutants diffusion has a community structure, and some critical air monitoring stations have a greater influence on the transmission of pollutants, which should be paid more attention to the following air quality management.

关键词： air pollutants complex network community detection label propagation algorithm

来源：评论

学校读者我要写书评

暂无评论

A Stable and Distributed Community Detection algorithm Based on Maximal Cliques

A Stable and Distributed Community Detection Algorithm Based...

引用

IEEE International Conference on Systems, Man, and Cybernetics (SMC)

作者： Gui, Feng Ma, Yunlong Zhang, Feng Liu, Min Yin, Rong Shen, Weiming Tongji Univ Sch Elect & Informat Engn Shanghai 200092 Peoples R China Tongji Univ Key Lab Embedded Syst & Serv Shanghai 200092 Peoples R China

ISBN: (纸本)9781479986965

In the research area of community detection which aims at detecting some highly cohesive vertex subsets in social network, there mainly exist some problems, such as the algorithms with comparatively excellent quality of the final partitioning usually have high time complexity and some other fast algorithms often result in low quality of partitioning or other disadvantages. Nowadays, the increasing demands for community detection in large-scale social networks necessitate the use of distributed and scalable methods to detect communities in an effective and efficient manner. label propagation algorithm (LPA), whose time complexity is O(m) on a network with m edges, is a near linear time algorithm to detect community effectively. Besides, owing to having good scalability, the parallel version of LPA (DLPA) is suitable for community detection in large-scale social networks. However, DLPA synchronously updates the vertices labels, which usually brings about label oscillations and results in low quality of partitioning. In this paper, we analyze the drawbacks of DLPA and propose a novel method C-DLPA, which combines DLPA with the notion of maximal cliques and at the same time utilizes a new updating mechanism that updating each node' label by probability of its adjacent nodes, to make final partitioning become more accurate and to avoid oscillations effectively. The experimental results show that C-DLPA has better performance is not only low time cost by as much to avoid oscillations but its community detection accuracy compared with DLPA.

关键词： community detection label propagation algorithm maximal cliques synchronous distributed algorithm

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：