检索结果-内蒙古大学图书馆

Improving Sharpness-Aware Minimization Using Label Smoothing and Adaptive Adversarial Cross-Entropy Loss

IEEE Access 2025年 13卷 100326-100337页

作者： Tanapat Ratchatorn Masayuki Tanaka Department of Systems and Control Engineering School of Engineering Institute of Science Tokyo Meguro City Tokyo Japan

Recent advances in learning algorithms have identified loss surface sharpness as an effective metric for reducing the generalization gap. Building on this principle, Sharpness-Aware Minimization (SAM) was introduced to improve model generalization and has achieved state-of-the-art performance through its weight perturbation and updating steps. However, SAM’s perturbation is based solely on the gradient of the standard cross-entropy loss. As the model approaches convergence, this gradient diminishes and oscillates, leading to inconsistent perturbation directions. To address these limitations, we introduce SAM-CE-AACE and SAM-SCE-AACE. SAM-CE-AACE replaces the conventional cross-entropy loss with the Adaptive Adversarial Cross-Entropy (AACE) loss in the perturbation step. Unlike the standard loss, AACE increases in both value and gradient magnitude as the model converges, ensuring stable perturbation directions. Additionally, SAM-CE-AACE’s perturbation utilizes AACE loss without normalization, enhancing the model’s exploratory capabilities in near-optimum stages. Moreover, SAM-SCE-AACE further improves SAM-CE-AACE by integrating label smoothing into the weight-updating step, reducing overconfidence in predictions. Together, this combination of the AACE loss function and label smoothing technique leads to further improvement in model performance. Intensive empirical experiments have been conducted to confirm the effectiveness of our methods, with experiments demonstrating improved performance in image classification tasks using different network architectures across various datasets. The reproduction code is available online at https://***/proj/AACE

关键词： Perturbation methods Training Standards Adaptation models Smoothing methods Minimization Vectors Probability distribution Feature extraction Market research

来源：评论

学校读者我要写书评

暂无评论

Site adaptation with machine learning for a Northern Europe gridded global solar irradiance product

引用

Energy and AI 2024年第1期15卷 265-278页

作者： Sebastian Zainali Dazhi Yang Tomas Landelius Pietro Elia Campana Mälardalen University Department of Sustainable Energy SystemsVästerås SE 72123Sweden School of Electrical Engineering and Automation Harbin Institute of TechnologyHarbinHeilongjiangChina Swedish Meteorological and Hydrological Institute Norrköping SE 60176Sweden

Gridded global horizontal irradiance(GHI)databases are fundamental for analysing solar energy applications’technical and economic aspects,particularly photovoltaic ***,there exist numerous gridded GHI databases whose quality has been thoroughly validated against ground-based irradiance ***,databases that generate data at latitudes above 65˚are few,and those available gridded irradiance products,which are either reanalysis or based on polar orbiters,such as ERA5,COSMO-REA6,or CM SAF CLARA-A2,generally have lower quality or a coarser time resolution than those gridded irradiance products based on geostationary *** the high-latitude gridded GHI databases,the STRÅNG model developed by the Swedish Meteorological and Hydrological institute(SMHI)is likely the most accurate one,providing data across *** further enhance the product quality,the calibration technique called"site adaptation"is herein used to improve the STRÅNG dataset,which seeks to adjust a long period of low-quality gridded irradiance estimates based on a short period of high-quality irradiance *** study introduces a novel approach for site adaptation of solar irradiance based on machine learning techniques,which differs from the conventional statistical methods used in previous *** machine-learning algorithms have been analysed and compared with conventional statistical approaches to identify Sweden’s most accurate algorithms for site *** irradiance data gathered from three weather stations of SMHI is used for training and *** results show that machine learning can substantially improve the STRÅNG model’s ***,due to the spatiotemporal heterogeneity in model performance,no universal machine learning model can be identified,which suggests that site adaptation is a location-dependant procedure.

关键词： Machine learning Global horizontal irradiance STRÅNG Site adaptation Agrivoltaic Sweden

来源：评论

学校读者我要写书评

暂无评论

The Journey/DAO/TAO of Embodied Intelligence: From Large Models to Foundation Intelligence and Parallel Intelligence

引用

IEEE/CAA Journal of Automatica Sinica 2024年第6期11卷 1313-1316页

作者： Tianyu Shen Jinlin Sun Shihan Kong Yutong Wang Juanjuan Li Xuan Li Fei-Yue Wang College of Information Science and Technology Beijing University of Chemical TechnologyBeijing 100029China School of Electrical and Information Engineering Jiangsu UniversityZhenjiang 212013China State Key Laboratory for Turbulence and Complex Systems Department of Advanced Manufacturing and RoboticsCollege of EngineeringPeking UniversityBeijing 100871China State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of AutomationChinese Academy of SciencesBeijing 100190China Peng Cheng Laboratory Shenzhen 518000China State Key Laboratory for Management and Control of Complex Systems Chinese Academy of SciencesBeijing 100190 Faculty of Innovation Engineering Macao University of Science and TechnologyMacao 999078China IEEE

THE tremendous impact of large models represented by ChatGPT[1]-[3]makes it necessary to con-sider the practical applications of such models[4].However,for an artificial intelligence(AI)to truly evolve,it needs to possess a physical“body”to transition from the virtual world to the real world and evolve through interaction with the real *** this context,“embodied intelligence”has sparked a new wave of research and technology,leading AI beyond the digital realm into a new paradigm that can actively act and perceive in a physical environment through tangible entities such as robots and automated devices[5].

关键词： Intelligence DAO TAO

来源：评论

学校读者我要写书评

暂无评论

Dynamic Blockchain-Empowered Trustworthy End-Edge Collaborative Computing via Rotating Multi-Agent DRL

引用

IEEE Transactions on Wireless Communications 2025年第6期24卷 4864-4878页

作者： Xu, Chi Zhang, Peifeng Yu, Haibin Li, Yonghui Chinese Academy of Sciences State Key Laboratory of Robotics Shenyang Institute of Automation Shenyang China Chinese Academy of Sciences Key Laboratory of Networked Control Systems Shenyang China University of Chinese Academy of Sciences Beijing China University of Sydney School of Electrical and Information Engineering Sydney Australia

Blockchain-empowered end-edge collaborative computing is a promising technology for enhancing the timeliness and trustworthiness of Industrial Internet of Things (IIoT). However, integrating task offloading with blockchain consensus inevitably escalates resource consumption across communication, computation, and energy domains. Thus, the joint optimization of task offloading, resource allocation and blockchain consensus is very important for IIoT. This paper studies a general end-edge collaborative computing scenario with multiple end devices and multiple edge servers. We first propose a novel dynamic blockchain (DBC) scheme by developing a dynamic leader election mechanism and designing a dynamic consensus waiting time window. Then, by fully considering the constraints of multi-task size and deadline, communication bandwidth, computing frequency, battery capacity, Byzantine fault tolerant and trustworthiness, we formulate the trustworthy processing efficiency (TPE) maximization problem with respect to end-edge task division, communication and computation resource allocation, leader election and consensus waiting window. To address this problem, we transform it into a Markov decision process and design a compound reward by fully considering the penalty for computing timeout and consensus failure. After that, we propose a rotating multi-agent deep reinforcement learning (R-MADRL) algorithm tailored to the proposed DBC scheme, where an entropy-based dual-critic DRL algorithm is proposed for rotating multi-agent training and decentralized execution. Extensive experiments validate the effectiveness and superiority of the proposed DBC with R-MADRL, where three benchmark DRL algorithms and three blockchain consensus schemes are compared. The results demonstrate that R-MADRL achieves stable convergence with more than 60.32% TPE reward than other algorithms while the task timeout ratio of DBC is reduced by more than 66.49% compared with other schemes. © 2025 institute of El

关键词： Markov processes

来源：评论

学校读者我要写书评

暂无评论

Sliding Mode control for Recurrent Neural Networks With Time-Varying Delays and Impulsive Effects

引用

IEEE/CAA Journal of Automatica Sinica 2023年第5期10卷 1319-1321页

作者： Lanfeng Hua Kaibo Shi Zheng-Guang Wu Soohee Han Shouming Zhong the School of Automation Engineering University of Electronic Science and Technology of ChinaChengdu 611731China the School of Electronic Information and Electrical Engineering Chengdu UniversityChengdu 610106China the Institute of Cyber-Systems and Control Zhejiang UniversityHangzhou 310027China the Department of Electrical Engineering and Convergence IT Engineering Pohang University of Science and TechnologyPohang 37673South Korea the School of Mathematics Sciences University of Electronic Science and Technology of ChinaChengdu 611731China

Dear Editor,This letter studies the problem of sliding mode control(SMC)design for recurrent neural networks(RNNs)with impulsive disturbances and time-varying transmission *** this end,an appropriate integral sliding ... 详细信息

关键词： impulsive letter delays

来源：评论

学校读者我要写书评

暂无评论

Grasp Detection with Hierarchical Multi-Scale Feature Fusion and Inverted Shuffle Residual

引用

Tsinghua Science and Technology 2024年第1期29卷 244-256页

作者： Wenjie Geng Zhiqiang Cao Peiyu Guan Fengshui Jing Min Tan Junzhi Yu State Key Laboratory of Multimodal Artificial Intelligence Systems Institute of AutomationChinese Academy of SciencesBeijing 100190China School of Artificial Intelligence University of Chinese Academy of SciencesBeijing 100049China Department of Advanced Manufacturing and Robotics College of EngineeringPeking UniversityBeijing 100871China

Grasp detection plays a critical role for robot *** pixel-wise grasp detection networks with encoder-decoder structure receive much attention due to good accuracy and ***,they usually transmit the high-level feature in the encoder to the decoder,and low-level features are *** is noted that low-level features contain abundant detail information,and how to fully exploit low-level features remains ***,the channel information in high-level feature is also not well ***,the performance of grasp detection is *** solve these problems,we propose a grasp detection network with hierarchical multi-scale feature fusion and inverted shuffle *** low-level and high-level features in the encoder are firstly fused by the designed skip connections with attention module,and the fused information is then propagated to corresponding layers of the decoder for in-depth feature *** a hierarchical fusion guarantees the quality of grasp ***,an inverted shuffle residual module is created,where the high-level feature from encoder is split in channel and the resultant split features are processed in their respective *** such differentiation processing,more high-dimensional channel information is kept,which enhances the representation ability of the ***,an information enhancement module is added before the encoder to reinforce input *** proposed method attains 98.9%and 97.8%in image-wise and object-wise accuracy on the Cornell grasping dataset,respectively,and the experimental results verify the effectiveness of the method.

关键词： grasp detection hierarchical multi-scale feature fusion skip connections with attention inverted shuffle residual

来源：评论

学校读者我要写书评

暂无评论

Response probability density of a single-degree-of-freedom linear system under non-Gaussian random excitation with dominant frequency 14

Response probability density of a single-degree-of-freedom l...

引用

14th International Conference on Recent Advances in Structural Dynamics, RASD 2024

作者： Tsuchida, Takahiro Department of Systems and Control Engineering Tokyo Institute of Technology Ookayama 2-12-1 Meguro-ku Tokyo152-8552 Japan

The response probability density function (PDF) of a single-degree-of-freedom linear system under non-Gaussian random excitation with a non-zero dominant frequency is investigated. The non-Gaussian excitation is a zero-mean stationary stochastic process prescribed by a first-order PDF and a power spectral density (PSD) with bandwidth and dominant frequency parameters. Uniform and Laplace distributions are considered for the excitation PDF. Monte Carlo simulations are performed to obtain the stationary displacement and velocity PDFs of the system. The simulation results indicate that the effects of excitation non-Gaussianity on the response PDFs vary greatly, depending on the bandwidth and dominant frequency of the excitation PSD. Particularly, the excitation non-Gaussianity appears strongly in the response PDFs when the excitation has a narrow bandwidth and the excitation dominant frequency matches the natural frequency of the system. Next, it is shown that when the excitation and response waveforms are similar, the response PDFs take a shape similar to the excitation PDF;on the other hand, when the waveforms are different, the response PDFs are nearly Gaussian, irrespective of the shape of the excitation PDF. Furthermore, as a quantitative index of waveform similarity, the maximum absolute value of the normalized cross-correlation function of excitation and response is introduced. The correspondence between the index value and the non-Gaussianity of the response PDF is revealed. © 2024 institute of Physics Publishing. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

A Day-Ahead Market Clearing Model Under System Inertia Constraints 16

A Day-Ahead Market Clearing Model Under System Inertia Const...

引用

16th IEEE PES Asia-Pacific Power and Energy engineering Conference, APPEEC 2024

作者： Ma, Yuting Jin, Jiangliang Hao, Liangliang Huang, Qisheng Shen, Liye College of Information Scienc and Technology Donghua University Shanghai China Shandong University Department of Control Science and Engineering Shandong China School of Mechanical Engineering and Automation Harbin Institute of Technology Shanghai China

ISBN: (纸本)9798350386127

We study the market clearing model for electric energy, inertia, and reserve in the day-ahead market, with a particular focus on the virtual inertia provided by wind units. In this paper, we consider synchronous inertia and virtual inertia as sources of system inertia. We incorporate system inertia constraints into the unit commitment (UC) and the economic dispatch (ED) models, and assign economic value to virtual inertia, so as to incentivize wind power to provide virtual inertia. Numerical results demonstrate that the proposed model enhances the participation of wind power units, increases the wind utilization rate and reduces the total costs. © 2024 IEEE.

关键词： Wind power

来源：评论

学校读者我要写书评

暂无评论

Stabilization With Prescribed Instant for High-Order Integrator systems

引用

IEEE Transactions on Cybernetics 2023年第11期53卷 7275-7284页

作者： Kuang, Jiyuan Gao, Yabin Chen, Chih-Chiang Zhang, Xiaoju Sun, Yizhuo Liu, Jianxing Harbin Institute of Technology Department of Control Science and Engineering Harbin150001 China National Cheng Kung University Department of Systems and Naval Mechatronic Engineering Tainan70101 Taiwan Harbin Institute of Technology School of Mathematics Harbin150001 China

This article develops a new controller design approach to stabilize system states onto the equilibrium at an arbitrarily selected time instant irrespective of the initial system states and parameters. By the stabilization approach, the actual convergence time (not the bound of actual convergence time) is independent of the initial value of system states. This feature differentiates our proposed prescribed-instant stability from conventional fixed, predefined, and prescribed time stability. In this work, we propose the controller design method for the prescribed-instant stability of n-order integrator systems. The proposed control is bounded and can gradually go to zero at an arbitrarily selected time instant, at which the system states reach zero simultaneously. This special stability of the controlled system is analyzed by reduction to absurdity. In simulations, an example of comparison with frequently used prescribed-time control is presented to show the difference. Moreover, the proposed stabilization method is validated by a magnetic suspension system with matched disturbances. © 2013 IEEE.

关键词： Stability criteria

来源：评论

学校读者我要写书评

暂无评论

A Three-Dimensional Force Estimation Method for the Cable-Driven Soft Robot Based on Monocular Images

A Three-Dimensional Force Estimation Method for the Cable-Dr...

引用

2024 IEEE International Conference on Robotics and Biomimetics, ROBIO 2024

作者： Zhu, Xiaohan Bu, Ran Li, Zhen Xu, Fan Wang, Hesheng School of Electronic Information and Electrical Engineering Shanghai Jiao Tong University Department of Automation China Artificial Intelligence Research Institute School of Information and Control Engineering China University of Mining and Technology China

ISBN: (纸本)9781665481090

Soft manipulators are known for their superiority in coping with high-safety-demanding interaction tasks, e.g., robot-assisted surgeries, elderly caring, etc. Yet the challenges residing in real-time contact feedback have hindered further applications in precise manipulation. This paper proposes an end-to-end network to estimate the 3D contact force of the soft robot, with the aim of enhancing its capabilities in interactive tasks. The presented method features directly utilizing monocular images fused with multidimensional actuation information as the network inputs. This approach simplifies the preprocessing of raw data compared to related studies that utilize 3D shape information for network inputs, consequently reducing configuration reconstruction errors. The unified feature representation module is devised to elevate low-dimensional features from the system's actuation signals to the same level as image features, facilitating smoother integration of multimodal information. The proposed method has been experimentally validated in the soft robot testbed, achieving satisfying accuracy in 3D force estimation (with a mean relative error of 0.84% compared to the best-reported result of 2.2% in the related works). © 2024 IEEE.

关键词： Robotic surgery

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：