检索结果-内蒙古大学图书馆

State Derivative Normalization for Continuous-Time Deep Neural Networks

IFAC-PapersOnLine 2024年第15期58卷 253-258页

作者： Jonas Weigand Gerben I. Beintema Jonas Ulmen Daniel Görges Roland Tóth Maarten Schoukens Martin Ruskowski Chair of Machine Tools and Control Systems RPTU Kaiserslautern and the German Research Center for Artificial Intelligence Kaiserslautern Germany Control Systems (CS) Group at the Department of Electrical Engineering Eindhoven University of Technology Netherlands. R. Tóth is also affiliated to the Systems and Control Laboratory at the Institute for Computer Science and Control Budapest Hungary Institute for Electromobility RPTU Kaiserslautern Germany

The importance of proper data normalization for deep neural networks is well known. However, in continuous-time state-space model estimation, it has been observed that improper normalization of either the hidden state or hidden state derivative of the model estimate, or even of the time interval can lead to numerical and optimization challenges with deep learning based methods. This results in a reduced model quality. In this contribution, we show that these three normalization tasks are inherently coupled. Due to the existence of this coupling, we propose a solution to all three normalization challenges by introducing a normalization constant at the state derivative level. We show that the appropriate choice of the normalization constant is related to the dynamics of the to-be-identified system and we derive multiple methods of obtaining an effective normalization constant. We compare and discuss all the normalization strategies on a benchmark problem based on experimental data from a cascaded tanks system and compare our results with other methods of the identification literature.

关键词： System Identification Continuous-Time Neural Ordinary Differential Equations Nonlinear State-Space

来源：评论

学校读者我要写书评

暂无评论

Partially Occluded Face Expression Recognition with CBAM-Based Residual Network for Teaching Scene

Partially Occluded Face Expression Recognition with CBAM-Bas...

引用

2023 China automation Congress, CAC 2023

作者： Bai, Yanxing Chen, Luefeng Li, Min Wu, Min Pedrycz, Witold Hirota, Kaoru School of Automation China University of Geosciences Wuhan430074 China School of Automation Ministry of Education China University of Geosciences Hubei Key Laboratory of Advanced Control and Intelligent Automation for Complex Systems Engineering Research Center of Intelligent Technology for Geo-Exploration Wuhan430074 China University of Alberta Department of Electrical and Computer Engineering EdmontonABT6G 2R3 Canada Systems Research Institute Polish Academy of Sciences Warsaw00-901 Poland Istinye University Department of Computer Engineering Sariyer Istanbul34396 Turkey Tokyo Institute of Technology Yokohama226-8502 Japan

ISBN: (纸本)9798350303759

In this paper, a deep residual network based on convolutional block attention module (CBAM) is proposed, which is utilized for feature extraction of partially occluded face expression data. The proposed method overcomes the problem of localized occlusion face feature extraction by focusing on the regions and channels containing important information in the occluded face data through CBAM. Multi-task cascaded convolutional networks (MTCNN) are firstly utilized to localize the key regions of face emotion, and then deep emotion features are extracted by CBAM-ResNet network. The final emotion labels are generated. The effectiveness of this paper's method is verified on the RAF-DB dataset and the occluded CK+ dataset. The experimental accuracy in the RAF-DB dataset is 76.3%, which is 3.74% and 1.64% higher than the accuracy produced by the method of RGBT, and the WLS-RF, respectively. Application experiments are carried out in the real teaching scenario, which verifies the applicability of the algorithm in the real teaching scene. © 2023 IEEE.

关键词： Emotion Recognition

来源：评论

学校读者我要写书评

暂无评论

OPTIMAL DESIGN OF THERMAL MANAGEMENT SYSTEM FOR AIRBORNE DIRECTED ENERGY WEAPONS

OPTIMAL DESIGN OF THERMAL MANAGEMENT SYSTEM FOR AIRBORNE DIR...

引用

2024 CSAA/IET International Conference on Aircraft Utility systems, AUS 2024

作者： Liu, Yide Li, Yang Cheng, Chaoqian Zhao, Longfei School of Automation Science and Electrical Engineering Beihang University Beijing100191 China Ningbo Institute of Technology Beihang University Ningbo315800 China National Key Laboratory of Aircraft Integrated Flight Control Beijing100191 China Key Laboratory of Advanced Aircraft Utility Systems Beihang University Ministry of Industry and Information Technology China Tianmushan Laboratory Hangzhou311115 China

Next-generation aircraft will carry directed energy weapons to ensure air superiority and significantly enhance both defensive and offensive capabilities. However, these weapons face severe thermal management challenges. This paper contributes by finely analyzing the heating mechanisms of core components within directed energy weapons and proposes a thermal management solution based on pump-driven two-phase flow. The thermal flow of the chip array layout is modeled and analyzed. The cold plate structure is optimized for the best-balanced thermal resistance and pressure drop. © The Institution of Engineering & Technology 2024.

关键词： Directed-energy weapons

来源：评论

学校读者我要写书评

暂无评论

Skeleton-Based Multi-Stream Adaptive Graph Convolutional Network for Indoor Scene Action Recognition

Skeleton-Based Multi-Stream Adaptive Graph Convolutional Net...

引用

2023 China automation Congress, CAC 2023

作者： Li, Jiazhuo Chen, Luefeng Li, Min Wu, Min Pedrycz, Witold Hirota, Kaoru The School of Automation The Hubei Key Laboratory of Advanced Control and Intelligent Automation for Complex Systems The Engineering Research Center of Intelligent Technology for Geo-Exploration Ministry of Education China University of Geosciences Wuhan430074 China The Department of Electrical and Computer Engineering University of Alberta EdmontonABT6G 2R3 Canada The Systems Research Institute Polish Academy of Sciences Warsaw00-901 Poland The Department of Computer Engineering Istinye University Sariyer Istanbul34396 Turkey The Tokyo Institute of Technology Tokyo226-8502 Japan

ISBN: (纸本)9798350303759

With the rapid advances in computer vision, human action recognition has gradually received attention, but the current methods still exhibit some problems in indoor environments. The human skeleton, as the framework of human motion, contains high-quality actional feature information, and the skeleton-based action recognition method effectively avoid the interference of interior background noise and has advantages in indoor action recognition. The outstanding effect of graph convolutional networks on graph structure data processing has led to its rapid development and wide application in skeleton-based action recognition. Second-order skeletal information also contains a large number of actional features but is not effectively utilized. The artificial predefined topology of the human skeleton map has limitations, and cannot reflect the interaction between limbs. To solve the above problems, this article designs an adaptive weighted multi-stream graph convolutional network (AM-GCN) based on skeletal information, using an attention mechanism to enhance the network's ability to extract actional features, and an adaptive layer to make the construction graph more flexible, incorporating second-order skeletal features through a dual-stream architecture. In this article, the NTU-RGB+D dataset has been used for the experiments, the results show that the method in this article has good results. © 2023 IEEE.

关键词： Convolution

来源：评论

学校读者我要写书评

暂无评论

Convection-UNet: A Deep Convolutional Neural Network for Convection Detection based on the Geo High-speed Imager of Fengyun-4B

Convection-UNet: A Deep Convolutional Neural Network for Con...

引用

Pattern Recognition, Machine Vision and Intelligent Algorithms (PRMVIA), International Conference on

作者： Yufei Wang Baihua Xiao The State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences & University of Chinese Academy of Sciences Beijing China The State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Beijing China

Deep convection can cause a variety of severe weather conditions such as thunderstorms, strong winds, and heavy rainfall. Satellite observations provide all-weather and multi-directional observations, facilitating the timely detection of such weather systems, which is crucial to saving lives and property. However, previous methods based on channel feature extraction and threshold filtering did not make full use of information in satellite images, which led to limitations on such complex problems as strong convection detection. In this study, we propose a novel framework of a deep learning-based model Convection-UNet to detect convection. We use channel 4 to 7 of FY-4B GHI that we select according to the microphysical properties of convection as input and radar reflectivity as label. We combine the detailed training time and test time data augmentation strategies and build a deep neural network to automatically extract spatial context features and achieve end-to-end learning. Results show that the performance of our method far exceeds the previous channel extraction combined with threshold filtering methods such as BT and BTD at least 0.24 on Fi-measure. We also show that our channel selection and data augmentation strategies are of great significance to detect convection.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Spatially Dense Multi-Target Track Segment Association Algorithm

Spatially Dense Multi-Target Track Segment Association Algor...

引用

Chinese control Conference (CCC)

作者： Yan Yao Liping Yan Yuanqing Xia Key Laboratory of Intelligent Control and Decision of Complex Systems School of Automation Beijing Institute of Technology Beijing China

In the field of radar data processing, track interruption seriously affects target tracking, track fusion, and other tasks. The existing track segment association algorithms have low correlation accuracy in dense distributed or long-time interruption situations. To this purpose, a dense multi-target track segment association (DMTTSA) algorithm is proposed. Firstly, two identical networks based on the multi-head probability sparse (ProbSparse) self-attention are used to capture the long-term dependencies of the tracks. Then, the bidirectional quadruplet hard sample loss (BiQuaHard loss) is constructed to make the tracks belonging to the same targets closer and the tracks belonging to the different targets farther. Finally, DMTTSA takes the closest track pairs in the feature space as the associated tracks and divides the unassociated tracks into the birth and dead tracks in chronological order. Some comparative experiments are carried out to show the anti-noise performance of the DMTTSA, as well as the effectiveness of solving the problem of dense multi-target track interruption.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Industrial Internet for intelligent manufacturing:past, present, and future

引用

Frontiers of Information Technology & Electronic Engineering 2024年第9期25卷 1173-1192页

作者： Chi XU Haibin YU Xi JIN Changqing XIA Dong LI Peng ZENG State Key Laboratory of Robotics Shenyang Institute of AutomationChinese Academy of SciencesShenyang 110016China Key Laboratory of Networked Control Systems Chinese Academy of SciencesShenyang 110016China Institutes for Robotics and Intelligent Manufacturing Chinese Academy of SciencesShenyang 110169China University of Chinese Academy of Sciences Beijing 100049China

Industrial Internet, motivated by the deep integration of new-generation information and communication technology(ICT) and advanced manufacturing technology, will open up the production chain, value chain, and industry chain by establishing complete interconnections between humans, machines, and things. This will also help establish novel manufacturing and service modes, where personalized and customized production for differentiated services is a typical paradigm of future intelligent manufacturing. Thus, there is an urgent requirement to break through the existing chimney-like service mode provided by the hierarchical heterogeneous network architecture and establish a transparent channel for manufacturing and services using a flat network architecture. Starting from the basic concepts of process manufacturing and discrete manufacturing, we first analyze the basic requirements of typical manufacturing tasks. Then, with an overview on the developing process of industrial Internet, we systematically compare the current networking technologies and further analyze the problems of the present industrial *** this basis, we propose to establish a novel “thin waist” that integrates sensing, communication, computing, and control for the future industrial Internet. Furthermore, we perform a deep analysis and engage in a discussion on the key challenges and future research issues regarding the multi-dimensional collaborative sensing of task–resource, the end-to-end deterministic communication of heterogeneous networks, and virtual computing and operation control of industrial Internet.

关键词： Intelligent manufacturing Industrial Internet Thin waist Transparent service Manufacturing as a service

来源：评论

学校读者我要写书评

暂无评论

Learning Reduced-Order Linear Parameter-Varying Models of Nonlinear systems ⁎

引用

IFAC-PapersOnLine 2024年第15期58卷 265-270页

作者： Patrick J.W. Koelewijn Rajiv Singh Peter Seiler Roland Tóth Sioux Technologies B.V. Eindhoven The Netherlands Control Systems Group Eindhoven University of Technology Eindhoven The Netherlands The MathWorks Inc. Natick USA Department of Electrical Engineering and Computer Science University of Michigan Ann Arbor USA Systems and Control Laboratory HUN-REN Institute for Computer Science and Control Budapest Hungary

In this paper, we consider the learning of a Reduced-Order Linear Parameter-Varying Model (ROLPVM) of a nonlinear dynamical system based on data. This is achieved by a two-step procedure. In the first step, we learn a projection to a lower dimensional state-space. In step two, an LPV model is learned on the reduced-order state-space using a novel, efficient parameterization in terms of neural networks. The improved modeling accuracy of the method compared to an existing method is demonstrated by simulation examples.

关键词： Reduced-order modeling Neural Networks Linear Parameter-Varying Models

来源：评论

学校读者我要写书评

暂无评论

Fast Depth Estimation of Object via Neural Network Perspective Projection 11

Fast Depth Estimation of Object via Neural Network Perspecti...

引用

11th IEEE Data Driven control and Learning systems Conference, DDCLS 2022

作者： Han, Yu Chen, Yaran Li, Haoran Ma, Mingjun Zhao, Dongbin Institute of Automation Chinese Academy of Sciences State Key Laboratory of Management and Control for Complex Systems Beijing100190 China University of Chinese Academy of Sciences Beijing100000 China

ISBN: (数字)9781665496759

ISBN: (纸本)9781665496759

In autonomous driving and mobile robotic systems, obtaining the depths of objects in real-time is crucial. The current network-based methods usually design complex network to achieve 3D object detection or monocular depth estimation for the whole image, resulting in too slow to be applied to mobile robots. The perspective projection-based method can achieve real-time, which calculates the object depth based on the camera parameters, the object sizes in the world coordinates and in image coordinates. While it relies heavily on the accuracy of object size in images coordinates, and the size is usually obtained with errors through detector network. Combining the perspective projection-based methods and network-based methods, we propose a fast object depth estimation method by designing a neural network to learn perspective projection, called Fast-Depth-NPP: 1) Instead of considering the whole image, we only consider the local depth of the image;2) Using local image patches as network inputs avoids measurement errors of object size with detector;3) the use of global information is enhanced by incorporating position encoding. Our method is validated on the mobile robot public dataset Neurons Perception dataset, achieving excellent results and meeting the real-time requirements. © 2022 IEEE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Plate Shape Prediction Based on Data-Driven in Roll Quenching Process

Plate Shape Prediction Based on Data-Driven in Roll Quenchin...

引用

2023 China automation Congress, CAC 2023

作者： Hu, Liu Chen, Luefeng Hu, Jie Wu, Min Pedrycz, Witold Hirota, Kaoru School of Automation Ministry of Education China University of Geosciences Hubei Key Laboratory of Advanced Control and Intelligent Automation for Complex Systems The Engineering Research Center of Intelligent Technology for Geo-Exploration Wuhan430074 China University of Alberta Department of Electrical and Computer Engineering EdmontonABT6R 2G7 Canada Systems Research Institute Polish Academy of Sciences Warsaw00-901 Poland Istinye University The Department of Computer Engineering Istanbul Sariyer34396 Turkey Tokyo Institute of Technology Tokyo Yokohama226-8502 Japan

ISBN: (纸本)9798350303759

In the process of steel plate production, predicting the plate shape is of great significance for producing high-quality and consistently stable plate shapes. This paper presents a model that predicts both the defect types and flatness of the plate, providing theoretical support for setting process parameters in roller quenching production. First, the parameters of the quenching process are analyzed to identify their characteristics. Then, the K-Means clustering algorithm and correlation analysis are employed to process the quenching process parameters. A gradient boosting decision tree (GBDT) model is used to predict the defect types and flatness of the steel plates. Finally, industrial production data is utilized for experimental validation. The obtained experimental results verify the reliability of the proposed method. © 2023 IEEE.

关键词： Decision trees

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：