检索结果-内蒙古大学图书馆

27th European Conference on Artificial Intelligence, ECAI 2024

作者： Li, Weijie Gu, Zhentao Fan, Xiaochao Deng, Wenjun Yang, Yong Zhao, Xinyuan Diao, Yufeng Yang, Liang College of Computer Science and Technology Xinjiang Normal University China School of Software Xinjiang University China College of Computer Science and Technology Inner Mongolia Minzu University China School of Computer Science and Technology Dalian University of Technology China Key Laboratory of Social Computing and Cognitive Intelligence Ministry of Education Dalian China

ISBN: (纸本)9781643685489

The emergence of Large Language Models (LLMs) provides a new solution to text generation tasks that involve high complexity, such as text style transfer (TST) tasks. However, previous studies have not fully explored the TST capabilities of different LLMs, and have faced issues with a lack of uniform standards in the human evaluation stage. This makes the results of human evaluation difficult to reproduce and less credible. To address this, this paper designs a prompt template to guide the cutting-edge LLMs to perform effective text style transfer and carries out an in-depth comparative analysis of various small-scale language models. In the stage of human evaluation, this paper eschews the conventional rating system, opting instead for a comparative human assessment methodology, which we refer to as duel-ranking. This method determines the relative ranking of models through mutual comparison, serving as an alternative to direct scoring. Detailed evaluation instructions are provided herein, to enhance the reproducibility of this method and ensure consistency throughout the evaluation process. This manual evaluation process reveals that GPT-3.5 and GPT-4 exhibit excellent performance in the TST tasks. © 2024 The Authors.

关键词：

来源：评论

学校读者我要写书评

暂无评论

引用

25th International Conference on Web Information Systems Engineering, WISE 2024

作者： Tu, Longchuan Meng, Shunmei Liu, Xiao Liu, Guanfeng Beheshti, Amin Zhang, Xuyun School of Computer Science and Engineering Nanjing University of Science and Technology Nanjin China State Key Laboratory for Novel Software Technology Nanjing University Nanjin China School of Computing Macquarie University Sydney Australia

ISBN: (纸本)9789819605699

Recently, graph neural networks(GNNs) have played a key crucial in many recommendation situations. In particular, contrastive learning-based hypergraph neural networks (HGNNs) are gradually becoming a research focus for addressing issues of data sparsity and noise. Despite many studies proving their outstanding performance, there are still some shortcomings: i) Most contrastive learning-based HGNNs primarily rely on cross-view contrastive learning, while neglecting contrastive learning on the interaction graph. ii) Utilizing node embedding for hypergraph structure learning is susceptible to the influence of low-quality representations, thereby constraining the learning capability of HGNNs. To address these issues, we offer a Semantic Similarity-based Graph Contrastive Learning framework (SSGCL), which aims to jointly learn representations with rich semantic information through within-view and cross-view contrastive learning. Specifically, we first introduce consistency contrastive learning, which captures self-supervised signals through semantic similarity between nodes and their neighborhoods. It better captures the unique features of nodes through the connections between nodes and their neighborhoods. Then, we utilize interaction graph learning hypergraph structure, promoting it to extract potential node connections and thus improving its ability to describe homogeneous node relationships. Experimental evaluations on three actual datasets show that SSGCL performs much better than the current baseline models. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2025.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

PSA-CLNet: Pedestrian Search Method Based on Polarized Self-Attention and COIM Loss 9

PSA-CLNet: Pedestrian Search Method Based on Polarized Self-...

引用

9th International Conference on Virtual Reality, ICVR 2023

作者： Fu, Junjie Liu, Rui Yi, Pengfei Dong, Jing Zhou, Dongsheng Wei, Xiaopeng Dalian University National and Local Joint Engineering Laboratory of Computer Aided Design School of Software Engineering Dalian China Dalian University of Technology School of Computer Science and Technology Dalian China

ISBN: (纸本)9798350345810

Pedestrian search is a combined task of pedestrian detection and pedestrian re-identification, i.e., how to find a specific pedestrian in a panoramic picture. However, due to the impact of different imaging conditions and environments, the accuracy of pedestrian search is still low at present, which is difficult to meet the needs of practical applications. To improve the accuracy of pedestrian search, this paper proposes PSA-CLNet pedestrian search method, which integrates polarized self-attention mechanism (PSA) and cosine online instance matching (COIM) loss. Firstly, the channel attention is added to enhance the pedestrian feature information and improve the expression ability of the feature information. Secondly, spatial attention is added to enhance the areas that have more contribution to the pedestrian search task to increase the correlation between the pictures and improve the recognition ability. Meanwhile, this paper proposes a COIM loss based on the online instance matching (OIM) loss. The loss function can enlarge the class distance between pedestrians and background, and reduce the class distance between pedestrians. This could improve the accuracy of pedestrian search algorithm effectively. Finally, this paper uses ResNet as the main backbone network to verify the proposed method. The pedestrian search method proposed in this paper could achieve 96.60% of mAP index and 96.83% of top-1 index in CUHK-SYSU dataset, and achieve 48.78% of mAP index and 88.38% of top-1 index in PRW dataset. © 2023 IEEE.

关键词： COIM loss pedestrian search PSA

来源：评论

学校读者我要写书评

暂无评论

Improving Adaptive Runoff Forecasts in Data-Scarce Watersheds Through Personalized Federated Learning 27th

Improving Adaptive Runoff Forecasts in Data-Scarce Watershe...

引用

27th International Conference on Pattern Recognition, ICPR 2024

作者： Xie, Zaipeng Zhang, Xiangqin Wang, Yunfei Jie, Xuanyao Fang, Wenhao Cai, Yanping Key Laboratory of Water Big Data Technology of Ministry of Water Resources Hohai University Nanjing China College of Computer Science and Software Engineering Hohai University Nanjing China

ISBN: (纸本)9783031781827

Runoff forecasting plays a crucial role in water resource management and flood mitigation, but it often faces significant challenges due to data deficiency and decentralized datasets. Inadequate hydrological data in many watersheds hinders the development of accurate prediction models, while organizational barriers and concerns about data privacy lead to information silos, preventing effective collaborative modeling. To address these issues, we propose FedHydroDSW, a novel personalized federated learning framework for runoff forecasting in data-scarce watersheds. Our approach enables decentralized model training across multiple organizations without requiring direct data sharing, thus preserving data privacy. FedHydroDSW incorporates unique model similarity metrics and parameter adaptation methods, utilizing insights from data-rich areas to refine forecasts in data-scarce regions. Experiments on the CAMELS dataset show that FedHydroDSW significantly outperforms standalone and conventional federated learning methods, with an average 30.1% increase in NSE, and reductions of 28% in RAE and 26.6% in RMSE across multiple data-scarce basins. By enabling secure cross-silo learning, our FedHydroDSW strategy paves the way toward more equitable runoff forecasting globally. The customized training and pattern recognition techniques represent major progress in deploying federated intelligence for impactful hydrological predictions. Our code is publicly available at https://***/xyjie37/FedHydroDSW. © The Author(s), under exclusive license to Springer Nature Switzerland AG 2025.

关键词： Federated learning

来源：评论

学校读者我要写书评

暂无评论

Wiki2GH: A Recommendation Service to Link software Engineering Knowledge to Practical Development 17th

Wiki2GH: A Recommendation Service to Link Software Engineeri...

引用

CCF 17th International Conference on Service Science, CCF ICSS 2024

作者： Zhou, Yuqi Sun, Yanchun Wu, Jiawei Zhang, Jiaqi Huang, Gang Key Laboratory of High Confidence Software Technologies Ministry of Education Beijing China School of Computer Science Peking University Beijing China School of Software and Microelectronics Peking University Beijing China National Key Laboratory of Data Space Technology and System Beijing China

ISBN: (纸本)9789819757596

High-quality software applications are built on the knowledge and best practices of software engineering. However, there is a huge gap between the theoretical knowledge of software engineering and its practical implementation for software developers, especially beginners of software development. This paper proposes an innovative recommendation service named Wiki2GH to link Wikipedia software engineering knowledge to GitHub frameworks and libraries, aiming to help software developers systematically learn knowledge and apply it into practice. To implement the Wiki2GH recommendation service, this paper first uses the public data of Wikipedia to build a hierarchy software development knowledge catalog as the entrance of the service. Next, we propose a deep learning method to identify frameworks and libraries and construct a knowledge graph of open source software repositories based on dependency graph. Then, we propose a combined application recommendation service of frameworks and libraries in the development practice. The methods proposed in this paper are all evaluated, including the repository type classification method, the recommendation method, and the knowledge graph construction method. Experimental results show the effectiveness of the recommendation service proposed in this paper. © The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd. 2024.

关键词： Knowledge graph

来源：评论

学校读者我要写书评

暂无评论

Enrich Code Search Query Semantics with Raw Descriptions 19th

Enrich Code Search Query Semantics with Raw Descriptions

引用

19th EAI International Conference on Collaborative Computing: Networking, Applications and Worksharing, CollaborateCom 2023

作者： Liu, Xiangzheng Liu, Jianxun Hu, Haize Liu, Yi School of Computer Science and Engineering Hunan University of Science and Technology Hunan Xiangtan China Hunan Provincial Key Laboratory for Services Computing and Novel Software Technology Hunan University of Science and Technology Hunan Xiangtan China

ISBN: (纸本)9783031545207

Code search can recommend relevant source code according to the development intention (query statement) of the demander, thereby improving the efficiency of software development. In the research of deep code search model, code description is used to replace query sentences for training. However, the heterogeneity existing between the query statement and the code description will seriously affect the accuracy of the code search model. In order to make up for the shortcomings of code search, this paper proposes a sentence-integrated query expansion method—SIQE. Unlike previous query expansion methods that focus on word-level expansion, SIQE uses the entire code description fragment as the source of query expansion. And by learning the mapping relationship between the query statement and the code description, the heterogeneity problem between them is compensated. In order to verify the effect of the proposed model in code search tasks, the article conducts code search experiments and analyzes on two languages: python and java. Experimental results show that, compared with the baseline model, SIQE has higher code search results. Therefore, the SIQE model can effectively improve the search effect of query statements, improve the accuracy of code search, and further improve the development of software engineering. © ICST Institute for computer Sciences, Social Informatics and Telecommunications Engineering 2024.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

The internal circulations on internal mass transfer rate of a single drop in nonlinear uniaxial extensional flow

引用

Chinese Journal of Chemical Engineering 2023年第7期59卷 51-60页

作者： Anjun Liu Jie Chen Meng Guo Chengmin Chen Meihong Yang Chao Yang Qilu University of Technology(Shandong Academy of Sciences) Shandong Computer Science Center(National Supercomputer Center in Jinan)Jinan 250101China CAS Key Laboratory of Green Process and Engineering Institute of Process EngineeringChinese Academy of SciencesBeijing 100190China Jinan Key Laboratory of High Performance Industrial Software Jinan Institute of Supercomputer TechnologyJinan 251013China

The internal flow of a droplet in the nonlinear extensional flow field will exhibit more than two internal circulations with the variation of nonlinear intensity(E).In this paper,the effect of positions and sizes of internal circulations on internal mass transfer rate of a single spherical droplet in a nonlinear extensional flow field is studied and compared with that in a linear extensional flow *** simulation results show that when E≥0,there are two symmetrical internal circulations in the droplet,which is the same with that in a linear extensional *** limit value of mass transfer rate Sh is 15,which is equal to that in a linear extensional flow,no matter how large E *** E≤-3/7,the number of internal flow circulation of a droplet increase to four and the transfer rate Sh *** E=-1,the maximum internal transfer rate Sh equals 30 which is twice of that in a linear extensional *** generation of new flow circulations in droplets and the circulation positions will enhance mass transfer when E≤-3/7,which provides a new idea for enhancing the internal mass transfer rate of droplets.

关键词： Internal mass transfer Nonlinear extensional flow field Spherical droplet Internal circulations

来源：评论

学校读者我要写书评

暂无评论

Reward Maximization for Disaster Zone Monitoring with Heterogeneous UAVs

引用

IEEE/ACM Transactions on Networking 2024年第1期32卷 890-903页

作者： Xu, Wenzheng Wang, Chengxi Xie, Hongbin Liang, Weifa Dai, Haipeng Xu, Zichuan Wang, Ziming Guo, Bing Das, Sajal K. Sichuan University College of Computer Science Chengdu610065 China City University of Hong Kong Department of Computer Science Hong Kong Hong Kong Nanjing University Department of Computer Science and Technology Nanjing210023 China Dalian University of Technology School of Software Dalian116024 China Sichuan University Key Laboratory of Birth Defects and Related Maternal and Child Diseases West China Second Hospital College of Computer Science Chengdu610066 China Missouri University of Science and Technology Department of Computer Science RollaMO65409 United States

In this paper, we study the deployment of K heterogeneous UAVs to monitor Points of Interest (PoIs) in a disaster zone, where a PoI may represent a school building or an office building, in which people are trapped. A UAV can take images/videos of PoIs and send its collected information back to a nearby rescue station for decision-making. Unlike most existing studies that focused on only homogeneous UAVs, we here study the scheduling of K heterogeneous UAVs, where different UAVs have different energy capacities and functionalities that lead to different monitoring qualities (monitoring rewards) of each PoI. For example, one type of UAVs can take only visual images while the other type of UAVs can take both visual and thermal infrared images. In this paper, we investigate a problem of scheduling K heterogeneous UAVs to monitor PoIs so that the sum of monitoring rewards received by all UAVs is maximized, subject to energy capacity on each UAV. We propose the very first 13 -approximation algorithm for this scheduling problem. We also evaluate the performance of the proposed algorithm, using real parameters of commercial UAVs. Experimental results show that the performance of the proposed algorithm is promising, which is improved by 25%, compared with existing algorithms. © 1993-2012 IEEE.

关键词： Approximation algorithms

来源：评论

学校读者我要写书评

暂无评论

EMARec: a sequential recommendation with exponential moving average

引用

Neural Computing and Applications 2024年第21期36卷 12917-12933页

作者： Chen, Rui Wang, Zonglin Tang, Cundong Zhang, Jianwei Li, Pu Kong, Xiangjie Huang, Min College of Software Engineering Zhengzhou University of Light Industry Zhengzhou450001 China Henan Joint International Research Laboratory of Computer Animation Implementation Technologies Zhengzhou450001 China School of Information Science and Technology Northwest University Xi’an710127 China Faculty of Information and Engineering Xuchang Vocational Technical College Xuchang461000 China College of Computer Science and Technology Zhejiang University of Technology Hangzhou310023 China

Capturing dynamic preference features from user historical behavioral data is widely applied to improve the accuracy of recommendations in sequential recommendation tasks. However, existing deep neural network-based sequential recommendation methods often ignore the noise information in user behavioral data that influence recommendation effectiveness, making recommendation models sensitive to noisy data. Additionally, these deep learning-based recommendation models often require a large number of parameters to capture user behavior patterns, leading to higher time complexity and susceptibility to overfitting due to noise fluctuations. To address these issues, in this paper, we apply the moving average concept from the field of time-series analysis to sequential recommendation tasks. The approach smooths sequence data, removes noise, making the data more stable and reliable, increases the model’s insensitivity to noise information, and better understands the evolution of user interests and behavioral patterns. Specifically, in our experiments, the moving average algorithm effectively denoises sequential data with low time complexity. Therefore, we propose a Sequential Recommendation with Exponential Moving Average with MLP architecture that makes the model more competitive in terms of time complexity. Experimental results conducted on two datasets demonstrate that EMARec outperforms state-of-the-art sequential recommendation methods across various common evaluation metrics. © The Author(s), under exclusive licence to Springer-Verlag London Ltd., part of Springer Nature 2024.

关键词： Recommender systems

来源：评论

学校读者我要写书评

暂无评论

Deep visual-linguistic fusion network considering cross-modal inconsistency for rumor detection

引用

Science China(Information Sciences) 2023年第12期66卷 16-32页

作者： Yang YANG Ran BAO Weili GUO De-Chuan ZHAN Yilong YIN Jian YANG School of Computer Science and Engineering Nanjing University of Science and Technology National Key Laboratory for Novel Software Technology Nanjing University School of Software Shandong University

With the development of the Internet, users can freely publish posts on various social media platforms, which offers great convenience for keeping abreast of the world. However, posts usually carry many rumors, which require plenty of manpower for monitoring. Owing to the success of modern machine learning techniques, especially deep learning models, we tried to detect rumors as a classification problem automatically. Early attempts have always focused on building classifiers relying on image or text information, i.e., single modality in posts. Thereafter, several multimodal detection approaches employ an early or late fusion operator for aggregating multiple source information. Nevertheless, they only take advantage of multimodal embeddings for fusion and ignore another important detection factor, i.e., the intermodal inconsistency between modalities. To solve this problem, we develop a novel deep visual-linguistic fusion network(DVLFN) considering cross-modal inconsistency, which detects rumors by comprehensively considering modal aggregation and contrast information. Specifically, the DVLFN first utilizes visual and textual deep encoders, i.e., Faster R-CNN and bidirectional encoder representations from transformers, to extract global and regional embeddings for image and text modalities. Then, it predicts posts' authenticity from two aspects:(1) intermodal inconsistency, which employs the Wasserstein distance to efficiently measure the similarity between regional embeddings of different modalities, and(2) modal aggregation, which experimentally employs the early fusion to aggregate two modal embeddings for prediction. Consequently, the DVLFN can compose the final prediction based on the modal fusion and inconsistency measure. Experiments are conducted on three real-world multimedia rumor detection datasets collected from Reddit, Good News, and Weibo. The results validate the superior performance of the proposed DVLFN.

关键词： multimodal learning Wasserstein distance rumor detection

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：