检索结果-内蒙古大学图书馆

ENGINEERING RESEARCH EXPRESS 2025年第1期7卷

作者： Zhang, Qian Wei, Mingjun Liu, Jinyun Li, Hui Xu, Rongyang North China Univ Sci & Technol Coll Artificial Intelligence Tangshan 063210 Peoples R China Hebei Prov Key Lab Ind Intelligent Percept Tangshan 063210 Peoples R China

Federated learning (FL) is an emerging distributed machine learning paradigm that collaboratively trains a global model while preserving data privacy. However, in the presence of data heterogeneity, it encounters challenges such as slow convergence and reduced global model accuracy. To address this issue, we propose a novel personalized federated learning(PFL) algorithm, FedFP. First, the algorithm combines the global category embedding layer with a contrastive learning mechanism to address the inconsistency of feature distributions in heterogeneous data scenarios by sharing global category embedding information. Second, by incorporating personalized batch normalization and classifier layers, the algorithm mitigates the impact of data heterogeneity on model convergence and generalization, while better accommodating the personalized needs of clients. Finally, extensive experiments were conducted on widely used image classification datasets as well as real-world human activity recognition datasets. The experimental results show that FedFP outperforms existing federated learning algorithms.

关键词： federated learning heterogeneous data personalization feature extraction parameter decoupling

来源：评论

学校读者我要写书评

暂无评论

parameter decoupling Strategy for Semi-supervised 3D Left Atrium Segmentation 14

Parameter Decoupling Strategy for Semi-supervised 3D Left At...

引用

14th International Conference on Machine Vision (ICMV)

作者： Hao, Xuanting Gao, Shengbo Sheng, Lijie Zhang, Jicong Beihang Univ Sch Biol Sci & Med Engn Beijing Peoples R China Xidian Univ Sch Comp Sci & Technol Xian Shaanxi Peoples R China Beihang Univ Hefei Innovat Res Inst Hefei Anhui Peoples R China Beihang Univ Beijing Adv Innovat Ctr Biomed Engn Beijing Peoples R China Beihang Univ Beijing Adv Innovat Ctr Big Data Based Precis Med Beijing Peoples R China

ISBN: (数字)9781510650459

ISBN: (纸本)9781510650459;9781510650442

Consistency training has proven to be an advanced semi-supervised framework and achieved promising results in medical image segmentation tasks through enforcing an invariance of the predictions over different views of the inputs. However, with the iterative updating of model parameters, the models would tend to reach a coupled state and eventually lose the ability to exploit unlabeled data. To address the issue, we present a novel semi-supervised segmentation model based on parameter decoupling strategy to encourage consistent predictions from diverse views. Specifically, we first adopt a two-branch network to simultaneously produce predictions for each image. During the training process, we decouple the two prediction branch parameters by quadratic cosine distance to construct different views in latent space. Based on this, the feature extractor is constrained to encourage the consistency of probability maps generated by classifiers under diversified features. In the overall training process, the parameters of feature extractor and classifiers are updated alternately by consistency regularization operation and decoupling operation to gradually improve the generalization performance of the model. Our method has achieved a competitive result over the state-of-the-art semi-supervised methods on the Atrial Segmentation Challenge dataset, demonstrating the effectiveness of our framework. Code is available at https://***/BX0903/PDC.

关键词： Semi-supervised learning parameter decoupling Consistency Regularization Left Atrium Segmentation

来源：评论

学校读者我要写书评

暂无评论

Improved Direct Linear Transformation for parameter decoupling in Camera Calibration

引用

ALGORITHMS 2016年第2期9卷 31-31页

作者： Zhao, Zhenqing Ye, Dong Zhang, Xin Chen, Gang Zhang, Bin Harbin Inst Technol Sch Elect Engn & Automat Heilongjiang 150001 Peoples R China Capital Aerosp Machinery Co Beijing 100076 Peoples R China

For camera calibration based on direct linear transformation (DLT), the camera's intrinsic and extrinsic parameters are simultaneously calibrated, which may cause coupling errors in the parameters and affect the calibration parameter accuracy. In this paper, we propose an improved direct linear transformation (IDLT) algorithm for calibration parameter decoupling. This algorithm uses a linear relationship of calibration parameter errors and obtains calibration parameters by moving a three-dimensional template. Simulation experiments were conducted to compare the calibration accuracy of DLT and IDLT algorithms with image noise and distortion. The results show that the IDLT algorithm calibration parameters achieve higher accuracy because the algorithm removes the coupling errors.

关键词： camera calibration direct linear transformation parameter decoupling

来源：评论

学校读者我要写书评

暂无评论

A systematic method for Li-ion battery simplified electrochemical model parameter sensitivity ranking and optimization

引用

JOURNAL OF ENERGY STORAGE 2025年 109卷

作者： Liu, Enhui Sun, Leo Ziehl, Paul Univ South Carolina Coll Engn & Comp Columbia SC 29208 USA Dutch Fork High Sch Irmo SC 29063 USA

Electrochemical models are increasingly used for their ability to accurately describe the internal dynamics of lithium-ion batteries (LIBs). These models, however, involve many parameters, making their optimization both challenging and computationally expensive. To address this challenge, this paper proposes anew method for identifying and optimizing the parameters of a simplified electrochemical model (SEM) of LIBs. Unlike traditional parameter estimation and optimization techniques, this proposed method begins with a parameter sensitivity analysis to determine the identifiability of parameters, classifying and ranking their sensitivities by their influence on model accuracy. With the ranking of parameter sensitivities, parameter optimization can focus on those with high sensitivity, while those with low sensitivity can be assigned fixed values. To implement the proposed method, an SEM is first established, in which parameters with high sensitivity are prioritized for optimization, while those with low sensitivity are assigned fixed values. In implementation, high-sensitivity parameters are optimized using the particle swarm optimization (PSO) algorithm and fine-tuned with real-time data to enhance fidelity. The proposed method is validated with experimental data under a dynamic loading profile. Results demonstrate that this hybrid procedure can effectively and accurately identify SEM parameters. The proposed method provides systematic parameterization guidance for battery modeling. It incorporates the electrochemical perspective into parameterization, mitigates the impact of early parameter estimation steps on subsequent steps, and reduces the computational cost of optimization. The proposed method establishes a reliable foundation and systematic guidance for developing model-based battery management systems.

关键词： Lithium-ion battery (LIB) Simplified electrochemical model (SEM) parameter sensitivity parameter decoupling Particle swarm optimization (PSO)

来源：评论

学校读者我要写书评

暂无评论

Personalized federated few-shot node classification

引用

Science China(Information Sciences) 2025年第1期68卷 204-218页

作者： Yunfeng ZHAO Xintong HE Guoxian YU Jun WANG Yongqing ZHENG Carlotta DOMENICONI School of Software Shandong University Joint SDU-NTU Centre for Artificial Intelligence Research Shandong University Department of Mathematics National University of Singapore Department of Computer Science George Mason University

Personalized federated learning(PFL) aims to train customized models for individual clients in a decentralized setting, with the account of non-independent and identically distributed data across clients. However, most PFL methods adopt uniform classification layers for diverse clients and give rise to error-prone predictions, due to the task heterogeneity notably prominent in decentralized graph data scenarios. Although some PFL solutions setup client-specific classification layers for each client and optimize them only locally, they are corrupted with limited local training data. We propose an innovative solution called federated parameter decoupling and node augmentation(Fed PANo) to address these problems and to achieve personalized federated few-shot node classification, which is a prevalent and challenging but unexplored topic. Specifically, Fed PANo first separates the local model into the GNN and classifier to handle unique client-specific task variations. The GNN is trained through federated learning to capture shared knowledge of graph nodes across clients, while the classifier is custom-designed and trained individually for each client. Additionally, a generic classifier shared among clients is adopted to encourage the GNN's grasp of shared information. Then Fed PANo further proposes the node generator along with its local and collaborative training strategies to deal with the node scarcity of clients. Extensive experimental results on benchmark datasets confirm that Fed PANo outperforms eight competitive baselines across different settings.

关键词： personalized federated learning few-shot learning node classification task heterogeneity parameter decoupling node augmentation

来源：评论

学校读者我要写书评

暂无评论

Personalized Federated Class-Incremental Learning through Critical parameter Transfer

Personalized Federated Class-Incremental Learning through Cr...

引用

2025 IEEE International Conference on Acoustics, Speech, and Signal Processing, ICASSP 2025

作者： Wu, Feng Feng, Siwei Chen, Yuanlu Zhao, Libang School of Computer Science and Technology Soochow University SuZhou China

ISBN: (纸本)9798350368741

Personalized federated learning is a decentralized approach that enables clients to collaboratively train a shared model while customizing it to their unique data and requirements. However, existing methods are not suited for dynamic scenarios where clients continuously learn from new classes of data, as they struggle with the increased data heterogeneity caused by this influx. To address this, we propose a Personalized Federated Class Incremental Learning (PFCIL) framework based on critical parameter transfer, called CPT-PFCIL, which each local client can continuously learn new classes, and before uploading the local model to the server, the client separates parameters of the local model into global federated parameters and task-specific parameters. Through communication with the server, each client selectively receives knowledge from other clients, achieving continuous learning and updating of the model while providing personalized model optimization. Our method has been validated on three datasets, demonstrating better performance compared to existing personalized federated learning methods. © 2025 IEEE.

关键词： Critical parameter Transfer Data Heterogeneity parameter decoupling Personalized Federated Class-Incremental Learning

来源：评论

学校读者我要写书评

暂无评论

Current Prediction Error Based parameter Identification Method for SPMSM With Deadbeat Predictive Current Control

引用

IEEE TRANSACTIONS ON ENERGY CONVERSION 2021年第3期36卷 1700-1710页

作者： Zhou, Ying Zhang, Shuo Zhang, Chengning Li, Xueping Li, Xuerong Yuan, Xin Beijing Inst Technol Sch Mech Engn Beijing 100081 Peoples R China Beijing Inst Technol Collaborat Innovat Ctr Elect Vehicles Beijing Beijing 100081 Peoples R China

Deadbeat predictive current control (DPCC) can predict motor behavior based on SPMSM model. However, during the operation of motor system, motor parameters (such as stator inductance and flux linkage) vary frequently according to different working conditions, which may lead to controller parameter mismatch, causing current harmonic content to increase and efficiency to decrease. In order to solve these problems caused by parameter variation, first, this paper proposes a current prediction error model by considering uncertainties of model parameters. Second, stator inductance and flux linkage are decoupled based on current prediction error model, which can reduce the interaction between parameters. Finally, the Kalman Filter (KF) algorithm is presented to filter the decoupled parameters. It is shown that the stator inductance and flux linkage can be identified accurately and the complexity of computation can be simplified. The traditional DPCC method, Extended Kalman Filter (EKF) based DPCC method and the proposed DPCC method are comparatively analyzed in this paper. Simulation and experiment indicate that the proposed parameter decoupling identification method can effectively reduce current harmonic content, current fluctuation and current tracking errors caused by parameter mismatch.

关键词： Stators Inductance Couplings Predictive models Resistance Mathematical model Torque Surface-mounted permanent magnet synchronous machine (SPMSM) current prediction error model Kalman Filter (KF) parameter decoupling parameter identification

来源：评论

学校读者我要写书评

暂无评论

Step-by-Step Longitudinal Aerodynamic parameter Estimation Method for Fly-by-Wire Aircraft 11

Step-by-Step Longitudinal Aerodynamic Parameter Estimation M...

引用

11th International Conference on Mechanical and Aerospace Engineering (ICMAE)

作者： Zhao, Rong Wang, Lixin Wang, Lei Yue, Ting Beihang Univ Sch Aeronaut Sci & Engn Beijing Peoples R China

ISBN: (纸本)9781728183220

The fidelity of wind tunnel data often cannot meet the fidelity requirement of engineering simulator. Based on parameter decoupling principle, a step-by-step longitudinal parameter estimation method and a flight test procedure aiming at fly-by-wire aircraft was designed to avoid the disturb to estimation coming from the flight control system and decrease the collinearity of the flight data, as a result, improving the fidelity of estimation. The result of the parameter estimation reveals the response of dynamic model established on the basis of estimated parameter satisfied the tolerance of CCAR60, which means it meets the fidelity requirement of the engineering simulator. It also proves that the longitudinal parameter estimation method and the flight test procedure are capable of gaining a sound estimation.

关键词： fly-by-wire flight test parameter decoupling parameter estimation aerodynamic model validation

来源：评论

学校读者我要写书评

暂无评论

Identification of Multiple parameters of the Bammann-Chiesa-Johnson Constitutive Model with Comprehensive Experiments for Pure Aluminum

引用

JOURNAL OF MATERIALS ENGINEERING AND PERFORMANCE 2020年第2期29卷 1262-1271页

作者： Zhou, Tingting Wang, Gang Yang, Yang Li, Yao Shuai, Maobing Tsinghua Univ Beijing Key Lab Precis Ultraprecis Mfg Equipments Beijing 100084 Peoples R China China Acad Engn Phys Inst Mat Jiangyou 621908 Peoples R China Chengdu Aircraft Ind Grp Co Ltd AVIC Chengdu 610092 Peoples R China

The flow stress is significantly affected in the precision machining process by the factors of strain path, strain rate history, temperature history and even microstructures. The physics-based Bammann-Chiesa-Johnson viscoplastic model is capable of capturing the flow stress evolution as well as the rate and temperature dependence. However, it is extremely challenging to identify all 18 constants in the model. In this paper, a novel method was proposed to identify the material constants by decoupling the parameters of internal state variables. A comprehensive experimental plan was conducted to obtain the true stress-strain curves of aluminum 1060 at elevated temperatures from 25 to 500 degrees C and strain rates from 10(-6) to 16400 s(-1) using creeping, quasi-static and SHPB methods. The particle swarm optimization algorithm was adapted to find the optimal solutions of the model parameters. The results showed that the prediction on the established model matched well with the experimental data. The quantitative error analysis confirmed the reliability of the method across a large range of strain rate and temperature variations.

关键词： Aluminum 1060 BCJ constitutive model parameter decoupling parameter identification the particle swarm optimization

来源：评论

学校读者我要写书评

暂无评论

New approach for point pollution source identification in rivers based on the backward probability method

引用

ENVIRONMENTAL POLLUTION 2018年 241卷 759-774页

作者： Wang, Jiabiao Zhao, Jianshi Lei, Xiaohui Wang, Hao Tsinghua Univ Dept Hydraul Engn State Key Lab Hydrosci & Engn Beijing 100084 Peoples R China China Inst Water Resources & Hydropower Res Beijing 100038 Peoples R China

Pollution risk from the discharge of industrial waste or accidental spills during transportation poses a considerable threat to the security of rivers. The ability to quickly identify the pollution source is extremely important to enable emergency disposal of pollutants. This study proposes a new approach for point source identification of sudden water pollution in rivers, which aims to determine where (source location), when (release time) and how much pollutant (released mass) was introduced into the river. Based on the backward probability method (BPM) and the linear regression model (LR), the proposed LR-BPM converts the ill-posed problem of source identification into an optimization model, which is solved using a Differential Evolution Algorithm (DEA). The decoupled parameters of released mass are not dependent on prior information, which improves the identification efficiency. A hypothetical case study with a different number of pollution sources was conducted to test the proposed approach, and the largest relative errors for identified location, release time, and released mass in all tests were not greater than 10%. Uncertainty in the LR BPM is mainly due to a problem with model equifinality, but averaging the results of repeated tests greatly reduces errors. Furthermore, increasing the gauging sections further improves identification results. A real-world case study examines the applicability of the LR BPM in practice, where it is demonstrated to be more accurate and time-saving than two existing approaches, Bayesian MCMC and basic DEA. (C) 2018 Elsevier Ltd. All rights reserved.

关键词： Backward probability method Linear regression Multi-point pollution source identification parameter decoupling River

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：