检索结果-内蒙古大学图书馆

arXiv 2025年

作者： Fang, Sheng Liu, Yong-Jin Yao, Wei Yu, Chengming Zhang, Jin Public Health School Fujian Medical University China School of Mathematics and Statistics Fuzhou University China Center for Applied Mathematics of Fujian Province China National Center for Applied Mathematics Shenzhen SUSTech China Department of Mathematics SUSTech China School of Science BUPT China CETC Key Laboratory of Smart City Modeling Simulation and Intelligent Technology The Smart City Research Institute of CETC China

Bilevel optimization, addressing challenges in hierarchical learning tasks, has gained significant interest in machine learning. The practical implementation of the gradient descent method to bilevel optimization encounters computational hurdles, notably the computation of the exact lower-level solution and the inverse Hessian of the lower-level objective. Although these two aspects are inherently connected, existing methods typically handle them separately by solving the lower-level problem and a linear system for the inverse Hessian-vector product. In this paper, we introduce a general framework to address these computational challenges in a coordinated manner. Specifically, we leverage quasi-Newton algorithms to accelerate the resolution of the lower-level problem while efficiently approximating the inverse Hessian-vector product. Furthermore, by exploiting the superlinear convergence properties of BFGS, we establish the non-asymptotic convergence analysis of the BFGS adaptation within our framework. Numerical experiments demonstrate the comparable or superior performance of the proposed algorithms in real-world learning tasks, including hyperparameter optimization, data hyper-cleaning, and few-shot meta-learning. Copyright © 2025, The Authors. All rights reserved.

关键词： Linear systems

来源：评论

学校读者我要写书评

暂无评论

Self-Supervised Entity Alignment Based on Multi-Modal Contrastive Learning

引用

IEEE/CAA Journal of Automatica Sinica 2022年第11期9卷 2031-2033页

作者： Bo Liu Ruoyi Song Yuejia Xiang Junbo Du Weijian Ruan Jinhui Hu the School of Software Engineering Xi’an Jiaotong UniversityXi’an 710000 the China Electronics Technology Group Corporation(CETC)Key Laboratory of Smart City Model Simulation and Intelligent Technology the Smart City Research Institute of CETC and National Center for Applied Mathematics Shenzhen(NCAMS)Shenzhen 518000China the Chinese University of Hong Kong Shenzhen 518000China Tencent Shenzhen 518000China the CETC Key Laboratory of Smart City Model Simulation and Intelligent Technology the Smart City Research Institute of CETC and National Center for Applied Mathematics Shenzhen(NCAMS)Shenzhen 518000China the Shenzhen Institute of Advanced Technology Chinese Academy of SciencesShenzhen 518055

Dear Editor,This letter proposes an unsupervised entity alignment method,which realizes integration of multiple multi-modal knowledge graphs *** recent years,Large-scale multi-modal knowledge graphs(LMKGs),containing text and image,have been widely applied in numerous knowledge-driven topics,such as question answering,entity linking,information extraction,reasoning and recommendation.

关键词： modal linking letter

来源：评论

学校读者我要写书评

暂无评论

CONSTRAINED BI-LEVEL OPTIMIZATION: PROXIMAL LAGRANGIAN VALUE FUNCTION APPROACH AND HESSIAN-FREE ALGORITHM 12

CONSTRAINED BI-LEVEL OPTIMIZATION: PROXIMAL LAGRANGIAN VALUE...

引用

12th International Conference on Learning Representations, ICLR 2024

作者： Yao, Wei Yu, Chengming Zeng, Shangzhi Zhang, Jin National Center for Applied Mathematics Shenzhen SUSTech China Department of Mathematics SUSTech China Department of Mathematics and Statistics UVic Canada CETC Key Laboratory of Smart City Modeling Simulation and Intelligent Technology The Smart City Research Institute of CETC China

This paper presents a new approach and algorithm for solving a class of constrained Bi-Level Optimization (BLO) problems in which the lower-level problem involves constraints coupling both upper-level and lower-level variables. Such problems have recently gained significant attention due to their broad applicability in machine learning. However, conventional gradient-based methods unavoidably rely on computationally intensive calculations related to the Hessian matrix. To address this challenge, we devise a smooth proximal Lagrangian value function to handle the constrained lower-level problem. Utilizing this construct, we introduce a single-level reformulation for constrained BLOs that transforms the original BLO problem into an equivalent optimization problem with smooth constraints. Enabled by this reformulation, we develop a Hessian-free gradient-based algorithm-termed proximal Lagrangian Value function-based Hessian-free Bilevel Algorithm (LV-HBA)-that is straightforward to implement in a single loop manner. Consequently, LV-HBA is especially well-suited for machine learning applications. Furthermore, we offer non-asymptotic convergence analysis for LV-HBA, eliminating the need for traditional strong convexity assumptions for the lower-level problem while also being capable of accommodating non-singleton scenarios. Empirical results substantiate the algorithm's superior practical performance. © 2024 12th International Conference on Learning Representations, ICLR 2024. All rights reserved.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

A Modified Stochastic User Equilibrium Based Back-Propagation Method of Transportation Network State Estimation

A Modified Stochastic User Equilibrium Based Back-Propagatio...

引用

2020 IEEE International Conference on Advances in Electrical Engineering and Computer Applications, AEECA 2020

作者： Yan, Chenyang Yang, Jiawen Xu, Jin Wu, Guangzhou Lian, Liping Shenzhen Graduate School of Peking University School of Urban Planning and Design Shenzhen China Electronics Technology Group Corporation Postdoctoral Innovation Base of Smart City Research Institute of China Shenzhen China Smart City Research Institute of Cetc Cetc Key Laboratory of Smart City Modeling Simulation and Intelligent Technology Shenzhen China

ISBN: (数字)9781728165219

ISBN: (纸本)9781728165202

In this research, we propose a modified stochastic user equilibrium based back-propagation method (MSUEBPM) to simultaneously estimate the transportation network state including an origin-destination (OD) demand matrix, link flow, and coefficients in the logit-model-based SUE. It is assumed that travelers in the same zone have similar route choice behavior. Multi-type data are required in this work, i.e. population, total demand of travelers, OD demand, link travel time and link flow. They can be derived from the residential trip survey, smartphone cellular signaling data, global position system, traffic sensors and so on. A back-propagation algorithm is applied to minimize a composite and non-convex objective function. The first-order partial derivatives of the objective function on the estimated coefficients are obtained. The proposed algorithm is tested in a simple network and a large network. the estimated results are good enough for engineering usage. However, in the large network, the calculation speed is much slower for the much more routes. © 2020 IEEE.

关键词： Backpropagation

来源：评论

学校读者我要写书评

暂无评论

A CT Image Denoising Method Based on Projection Domain Feature

arXiv

引用

arXiv 2024年

作者： Sun, Mengyu Xia, Dimeng Zhao, Shusen Zhang, Weibin He, Yaobin Southern University of Science and Technology Shenzhen518055 China CETC Key Laboratory of Smart City Modeling Simulation and Intelligent Technology The Smart City Research Institute of CETC Shenzhen518055 China

In order to improve image quality of projection in industrial applications, generally, a standard method is to increase the current or exposure time, which might cause overexposure of detector units in areas of thin objects or backgrounds. Increasing the projection sampling is a better method to address the issue, but it also leads to significant noise in the reconstructed image. This paper proposed a projection domain denoising algorithm based on the features of the projection domain for this case. This algorithm utilized the similarity of projections of neighboring veiws to reduce image noise quickly and effectively. The availability of the algorithm proposed in this work has been conducted by numerical simulation and practical data experiments. Copyright © 2024, The Authors. All rights reserved.

关键词： Digital elevation model

来源：评论

学校读者我要写书评

暂无评论

Time-dependent system reliability analysis of a high concrete-faced rockfill dam by integrating multi-source monitoring information

引用

Reliability Engineering and System Safety 2025年 262卷

作者： Peng, Ming Gan, Longhou Zhong, Qiming Yang, Ge Deng, Gang Cao, Zijun Shi, Zhenming Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education Tongji University Shanghai China Department of Geotechnical Engineering College of Civil Engineering Tongji University Shanghai China Department of Geotechnical Engineering Nanjing Hydraulic Research Institute Nanjing China Key Laboratory of Reservoir and Dam Safety of Ministry of Water Resources Nanjing China Huadong Engineering Corporation Limited Hangzhou China National Hydropower Station Dam Safety and Emergency Engineering Technology Center Hangzhou China China Institute of Water Resources and Hydropower Research State Key Laboratory of Simulation and Regulation of Water Cycle in River Basin Beijing China MOE Key Laboratory of High-Speed Railway Engineering Institute of Smart City and Intelligent Transportation Southwest Jiaotong University Chengdu China

This paper presents a time-dependent reliability analysis method for concrete-faced rockfill dams (CFRDs) by integrating multiple failure modes and multi-source monitoring data via Bayesian networks. Initially, two sub-Bayesian networks are constructed to fuse dam parameters, two related failure modes, and three types of monitoring data. Subsequently, the prior failure probabilities of the dam system for each period are calculated through the time-variant response relationships among network nodes. These response relationships introduce a time-variant term to quantify the effects of water level and creep. Finally, various types of monitoring data are utilized to update parameter distribution, resulting in the posterior failure probabilities. The proposed method is applied to 233-meter-high Shuibuya CFRD. The results indicate that Bayesian networks offer a more comprehensive and reliable assessment. Water level induces periodic variations in system reliability, while creep drives the long-term trend by increasing slabs' failure probabilities. The failure probabilities of dam system increase over the initial ten years and stabilize as creep converges. The slabs’ failure probabilities vary from location. Seepage failure probability is primarily dominated by the most critical slab. Utilizing multi-source monitoring data can reduce uncertainties, mitigate the interference of localized abnormal data, and identify potential failure locations. This approach supports enhanced dam safety management. © 2025 Elsevier Ltd

关键词： Bayesian networks Concrete-faced rockfill dams (CFRDs) Information fusion Multiple failure modes Time-dependent reliability

来源：评论

学校读者我要写书评

暂无评论

Constrained Bi-Level Optimization: Proximal Lagrangian Value function Approach and Hessian-free Algorithm

arXiv

引用

arXiv 2024年

作者： Yao, Wei Yu, Chengming Zeng, Shangzhi Zhang, Jin National Center for Applied Mathematics Shenzhen Department of Mathematics Southern University of Science and Technology Shenzhen China CETC Key Laboratory of Smart City Modeling Simulation and Intelligent Technology The Smart City Research Institute of CETC National Center for Applied Mathematics Shenzhen Shenzhen China Department of Mathematics and Statistics University of Victoria VictoriaBC Canada Department of Mathematics Southern University of Science and Technology National Center for Applied Mathematics Shenzhen Shenzhen China

This paper presents a new approach and algorithm for solving a class of constrained Bi-Level Optimization (BLO) problems in which the lower-level problem involves constraints coupling both upper-level and lower-level variables. Such problems have recently gained significant attention due to their broad applicability in machine learning. However, conventional gradient-based methods unavoidably rely on computationally intensive calculations related to the Hessian matrix. To address this challenge, we begin by devising a smooth proximal Lagrangian value function to handle the constrained lower-level problem. Utilizing this construct, we introduce a single-level reformulation for constrained BLOs that transforms the original BLO problem into an equivalent optimization problem with smooth constraints. Enabled by this reformulation, we develop a Hessian-free gradient-based algorithm-termed proximal Lagrangian Value function-based Hessian-free Bi-level Algorithm (LV-HBA)-that is straightforward to implement in a single loop manner. Consequently, LV-HBA is especially well-suited for machine learning applications. Furthermore, we offer non-asymptotic convergence analysis for LV-HBA, eliminating the need for traditional strong convexity assumptions for the lower-level problem while also being capable of accommodating non-singleton scenarios. Empirical results substantiate the algorithm's superior practical performance. © 2024, CC BY-NC-SA.

关键词： Machine learning

来源：评论

学校读者我要写书评

暂无评论

Technological Innovation as a Panacea to Rent-Seeking Behavior: How Can the Diffusion of Continuous Emissions Monitoring Systems in Thermal Power Enterprises Be Promoted Based on an Evolutionary Game Model?

SSRN

引用

SSRN 2024年

作者： Wang, Yang Li, Dezhi Wei, Yigang Wang, Wentao Department of Construction and Real Estate School of Civil Engineering Southeast University Nanjing211189 China China-Pakistan Belt and Road Joint Laboratory on Smart Disaster Prevention of Major Infrastructures Ministry of Science and Technology Nanjing211189 China Engineering Research Center of Building Equipment Energy and Environment Ministry of Education Nanjing211189 China School of Economics and Management Beihang University Beijing100191 China Beijing Key Laboratory of Emergency Support Simulation Technologies for City Operations Beijing100191 China Laboratory for Low-carbon Intelligent Governance Beihang University Beijing100191 China

Carbon data fraud caused by rent-seeking behavior (RSB) undermines the effectiveness of the Emissions Trading Scheme (ETS). The Continuous Emissions Monitoring System (CEMS) used for carbon monitoring is a powerful solution to this misconduct. This study aims to optimize regulatory measures for promoting CEMS and curbing RSB (PCaCR) to improve the robustness of ETS. A tripartite evolutionary game model was constructed based on strategic interactions among thermal power enterprises (TPEs), third-party verifiers (3PVs), and government. The Lyapunov indirect method was used to analyze evolutionary stabilization strategies (ESS). Numerical simulations were performed to assess the sensitivity of key factors. The empirical results indicate that, first, four scenarios of ESS for the stakeholders. The ideal ESS for the PCaCR is when TPEs apply CEMS for carbon monitoring, 3PVs impartially verify, and the government strictly supervises. Second, the initial strategy choices of stakeholders affect the time it takes for the game system to reach the ESS. The closer the initial strategy selection is to the ESS, the faster the evolution occurs. Third, factors with different sensitivities affects the evolutionary outcome and speed. Standard refinement, information disclosure, and reward & punishment mechanisms can regulate stakeholder decision-making. © 2024, The Authors. All rights reserved.

关键词： Carbon

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：