检索结果-内蒙古大学图书馆

7th International Symposium on Autonomous Systems, ISAS 2024

作者： Mao, Tong Zheng, Hongyuan Zhai, Xiangping Bryce Zhu, Jing Nanjing University of Aeronautics and Astronautics College of Computer Science and Technology Nanjing China Nanjing University of Aeronautics and Astronautics College of Automation Engineering Nanjing China

ISBN: (纸本)9798350363173

In recent years, active object tracking (AOT), which is a distinct technology from passive target tracking, has made significant advances in the field of artificial intelligence. AOT aims to actively track a target object by autonomously controlling the tracker's perspective and position when the target object leaves the current tracking scene. Previous research has demonstrated that trackers can be trained to perform active object tracking tasks using deep reinforcement learning. However, due to the partial observability and high-dimensional state space characteristics of active object tracking tasks, challenges arise during the training process, resulting in low sample efficiency and unstable model training, which increases the difficulty of model training. To address these challenges, this paper proposes a novel deep reinforcement learning framework based on constrained policy optimization, named CPO-AOT. This framework combines the asynchronous training mechanism with the concept of policy constraints. It not only accelerates training and reduces redundancy but also effectively constrains the magnitude of policy updates, ensuring that the agent does not make excessively aggressive policy changes. Additionally, for the active object tracking problem, CPO-AOT specifically designs a sparse reward function suitable for high-dimensional state spaces to guide the model's attention towards crucial states. Experimental results conducted in a three-dimensional virtual environment demonstrate that the adopted CPO-AOT framework enhances the model's performance in challenging environments with high-dimensional state spaces and partial observability. It effectively prevents the model from getting stuck in local optima, thereby strengthening the model's robustness. Compared to other methods, the proposed CPO-AOT proves more suitable for addressing active object tracking problems in partially observable environments. © 2024 IEEE.

关键词： Deep learning

来源：评论

学校读者我要写书评

暂无评论

SELF-ENSEMBLE PROTECTION: TRAINING CHECKPOINTS ARE GOOD DATA PROTECTORS 11

SELF-ENSEMBLE PROTECTION: TRAINING CHECKPOINTS ARE GOOD DATA...

引用

11th International Conference on Learning Representations, ICLR 2023

作者： Chen, Sizhe Yuan, Geng Cheng, Xinwen Gong, Yifan Qin, Minghai Wang, Yanzhi Huang, Xiaolin Department of Automation Shanghai Jiao Tong University China Department of Electrical and Computer Engineering Northeastern University United States

As data becomes increasingly vital, a company would be very cautious about releasing data, because the competitors could use it to train high-performance models, thereby posing a tremendous threat to the company's commercial competence. To prevent training good models on the data, we could add imperceptible perturbations to it. Since such perturbations aim at hurting the entire training process, they should reflect the vulnerability of DNN training, rather than that of a single model. Based on this new idea, we seek perturbed examples that are always unrecognized (never correctly classified) in training. In this paper, we uncover them by model checkpoints' gradients, forming the proposed self-ensemble protection (SEP), which is very effective because (1) learning on examples ignored during normal training tends to yield DNNs ignoring normal examples;(2) checkpoints' cross-model gradients are close to orthogonal, meaning that they are as diverse as DNNs with different architectures. That is, our amazing performance of ensemble only requires the computation of training one model. By extensive experiments with 9 baselines on 3 datasets and 5 architectures, SEP is verified to be a new state-of-the-art, e.g., our small ∞ = 2/255 perturbations reduce the accuracy of a CIFAR-10 ResNet18 from 94.56% to 14.68%, compared to 41.35% by the best-known method. Code is available at https://***/Sizhe-Chen/SEP. © 2023 11th International Conference on Learning Representations, ICLR 2023. All rights reserved.

关键词：

来源：评论

学校读者我要写书评

暂无评论

Fusion of Attention Mechanism with Shape-Adaptive Rotational Object Detection 4

Fusion of Attention Mechanism with Shape-Adaptive Rotational...

引用

4th International Conference on Big Data and Artificial Intelligence and Software engineering, ICBASE 2023

作者： Li, Zhenxing Guo, Jialiang Hao, Yuzhi Harbin Engineering University College of Computer Science and Technology Harbin China Southeast University School of Automation Nanjing China

ISBN: (纸本)9798350329490

This study introduces a novel attention mechanism for object detection in optical remote sensing images. In aerial imagery, many small objects rely on environmental cues for recognition. Their successful recognition largely depends on their surrounding environmental backdrop since the environmental information can provide crucial clues about the object's shape, orientation, and other features. Therefore, we firmly believe in the enormous potential of the Swin Transformer in capturing these prior knowledge. Our strategy, based on RepPoints and using Swin Transformer as a backbone, engages in sample selection through the fusion of Shape-Adaptive Selection and Shape-Adaptive Measurement strategies. This strategy fully exploits the shape information of the targets, dynamically optimizing the process of positive and negative sample matching. We have also designed a boundary center loss function to enhance localization precision and incorporated an orientation classification prediction head to explicitly predict target angles. Experiments on a private SAR dataset and DOTA benchmark validate the superiority of our approach in improving accuracy, efficiency and reducing model size compared to other methods. This research paves the way for new directions in Transformer-based rotational object detection. © 2023 IEEE.

关键词： Object detection

来源：评论

学校读者我要写书评

暂无评论

Thermal Runaway Prediction of Lithium-ion Batteries based on Fusion Model 6

Thermal Runaway Prediction of Lithium-ion Batteries based on...

引用

6th International Conference on Electrical engineering and Control Technologies, CEECT 2024

作者： Yang, Kang Sun, Yuxiang Zhang, Kaibo Xia, Fei Shanghai University of Electric Power College of Automation Engineering Shanghai China Shanghai University of Electric Power College of Computer Science and Technology Shanghai China

ISBN: (纸本)9798331528096

Implementing a proactive alert system to detect thermal runaway in lithium-ion batteries is essential for ensuring the safety and dependability of energy infrastructures. In this research, a new approach is introduced that integrates an electrothermal coupling model with deep learning techniques to detect thermal runaway in its early stages. The proposed frameworks incorporate Convolutional Neural Networks (CNN), Bidirectional Long Short-Term Memory (BiLSTM) systems, and sequence-to-sequence neural models augmented with attention mechanisms. First, an electrothermal coupling model is constructed to simulate internal battery temperature changes. Subsequently, deep learning algorithms are employed to estimate battery voltage and surface temperature metrics. Finally, an internal temperature prediction model is established based on these predicted values. Experimental validation shows that this method controls the internal temperature prediction error within 3.4° C. It effectively achieves thermal runaway early warning and significantly improves the operational safety of lithium-ion batteries. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Active Learning for High-Quality Sawtooth Wave Dataset Generation in Sleep Analysis 6

Active Learning for High-Quality Sawtooth Wave Dataset Gener...

引用

6th International Conference on Frontier Technologies of Information and computer, ICFTIC 2024

作者： Chen, Jinghuang Bao, Xiaoyu Ye, Yuntao Huang, Qiyun School of Automation Science and Engineering South China University of Technology Guangzhou China Research Center for Brain-Computer Interface Pazhou Laboratory Guangzhou China

ISBN: (纸本)9798331541750

Sawtooth Waves (STW) that occur during Rapid Eye Movement (REM) sleep are closely associated with memory consolidation and the regulation of brain functions. However, due to challenges in data acquisition and the high cost of manual annotation, there is currently an absence of a publicly massive STW annotated dataset. This study focuses on constructing a high-quality STW dataset, comprising 4912 samples from 200 participants, and developing classification algorithms for STW detection. We utilized three prevalent algorithms-Support Vector Machine (SVM), Multilayer Perceptron (MLP), and Convolutional Neural Network (CNN)-to assess the dataset's quality. The dataset achieved a maximum classification accuracy of 95.62% with the CNN, confirming the gold standard of STW dataset. Finally, we studied age and gender difference in physiological characteristics of STW, including average frequency of STW, peak-to-peak amplitude(PTP), root mean square(RMS) and spectrum power. This study created a public STW dataset and a STW detection model, laying the groundwork for analyzing the mechanisms of STW events. © 2024 IEEE.

关键词： Convolutional neural networks

来源：评论

学校读者我要写书评

暂无评论

Towards Better Accuracy-Efficiency Trade-Offs: Dynamic Activity Inference via Collaborative Learning From Various Width-Resolution Configurations

IEEE Transactions on Artificial Intelligence

引用

IEEE Transactions on Artificial Intelligence 2024年第12期5卷 6723-6738页

作者： Qin, Lutong Zhang, Lei Li, Chengrun Song, Chaoda Cheng, Dongzhou Wang, Shuoyuan Wu, Hao Song, Aiguo Nanjing Normal University School of Electrical and Automation Engineering Nanjing210023 China Shanghai200240 China Case Western Reserve University Department of Computer and Data Sciences ClevelandOH44106 United States University of Macau Department of Computer and Information Science Taipa999078 China Yunnan University School of Information Science and Engineering Kunming650500 China Southeast University Department of Instrument Science and Engineering Nanjing210096 China

Recently, deep neural networks have triumphed over a large variety of human activity recognition (HAR) applications on resource-constrained mobile devices. However, most existing works are static and ignore the fact that the computational budget usually changes drastically across various devices, which prevent real-world HAR deployment. It still remains a major challenge: how to adaptively and instantly tradeoff accuracy and latency at runtime for on-device activity inference using time series sensor data? To address this issue, this article introduces a new collaborative learning scheme by training a set of subnetworks executed at varying network widths when fueled with different sensor input resolutions as data augmentation, which can instantly switch on-the-fly at different width-resolution configurations for flexible and dynamic activity inference under varying resource budgets. Particularly, it offers a promising performance-boosting solution by utilizing self-distillation to transfer the unique knowledge among multiple width-resolution configuration, which can capture stronger feature representations for activity recognition. Extensive experiments and ablation studies on three public HAR benchmark datasets validate the effectiveness and efficiency of our approach. A real implementation is evaluated on a mobile device. This discovery opens up the possibility to directly access accuracy-latency spectrum of deep learning models in versatile real-world HAR deployments. © 2020 IEEE.

关键词： Budget control

来源：评论

学校读者我要写书评

暂无评论

Detecting Double Mixed Compressed Images Based on Quaternion Convolutional Neural Network

引用

Chinese Journal of Electronics 2024年第3期33卷 657-671页

作者： Hao WANG Jinwei WANG Xuelong HU Bingtao HU Qilin YIN Xiangyang LUO Bin MA Jinsheng SUN Department of Automation Nanjing University of Science and Technology Engineering Research Center of Digital Forensics Ministry of Education Department of Computer and Software Nanjing University of Information Science and Technology School of Management Nanjing University of Posts and Telecommunications Department of Computer Science and Engineering Guangdong Province Key Laboratory of Information Security Technology Sun Yat-sen University State Key Laboratory of Mathematical Engineering and Advanced Computing Qilu University of Technology

Detection of color images that have undergone double compression is a critical aspect of digital image *** the existence of various methods capable of detecting double Joint Photographic Experts Group(JPEG) compression,they are unable to address the issue of mixed double compression resulting from the use of different compression *** particular,the implementation of Joint Photographic Experts Group 2000(JPEG2000)as the secondary compression standard can result in a decline or complete loss of performance in existing *** tackle this challenge of JPEG+JPEG2000 compression,a detection method based on quaternion convolutional neural networks(QCNN) is *** QCNN processes the data as a quaternion,transforming the components of a traditional convolutional neural network(CNN) into a quaternion *** relationships between the color channels of the image are preserved,and the utilization of color information is ***,the method includes a feature conversion module that converts the extracted features into quaternion statistical features,thereby amplifying the evidence of double *** results indicate that the proposed QCNN-based method improves,on average,by 27% compared to existing methods in the detection of JPEG+JPEG2000 compression.

关键词： Color image forensics JPEG JPEG2000 Mixed double compression Quaternion convolutional neural network

来源：评论

学校读者我要写书评

暂无评论

State of Charge Estimation for Zinc-Bromine Flow Batteries by Improved Long Short-Term Memory Network and Kalman Filter

State of Charge Estimation for Zinc-Bromine Flow Batteries b...

引用

2024 CPSS and IEEE International Symposium on Energy Storage and Conversion, CPSS and ISESC 2024

作者： Peng, Kaiwen Chen, Bichen Wang, Liguo Sidorov, Denis Dreglea, Aliona Wang, Zongjie School of Electrical Engineering and Automation Harbin Institute of Technology Harbin China School of Electrical Engineering and Automation Northeast Agricultural University Harbin China Melentiev Energy Systems Institute Siberian Branch of the Russian Academy of Sciences Irkutsk Russia University of Connecticut Department of Electrical and Computer Engineering Mansfield United States

ISBN: (纸本)9798350380514

In order to improve the accuracy of estimating the state of charge (SOC) of zinc-bromine flow batteries (ZBFB) in the discharge stage and overcome the problems caused by sudden voltage changes, this paper proposes an algorithm for estimating and correcting the SOC. The core contribution is the proposal of an improved Long Short-Term Memory (LSTM) neural network algorithm, which considers the effect of the circulating pump speed on the battery voltage during the discharge stage, and establishes a neural network model for the battery. The estimation result of the improved LSTM network has a mean square error of 3.322, representing a 12% improvement in estimation accuracy compared to the traditional LSTM network. Additionally, Kalman filtering is employed for error mitigation to further enhance estimation accuracy. After applying the proposed method, the maximum error between the estimated SOC and the actual SOC is 1.91%, and the average error is 0.30%. This indicates that the proposed method improves the accuracy of SOC estimation through enhanced real-time LSTM network estimation and Kalman filtering error cancellation, particularly during voltage sudden change in the discharge of ZBFB. © 2024 IEEE.

关键词： Kalman filters

来源：评论

学校读者我要写书评

暂无评论

Neural-mechanism-driven image block encryption algorithm incorporating a hyperchaotic system and cloud model

引用

Chinese Physics B 2022年第4期31卷 175-190页

作者： Peng-Fei Fang Han Liu Cheng-Mao Wu Min Liu School of Automation and Information Engineering Xi’an University of TechnologyXi’an 710048China School of Electronic Engineering Xi’an University of Posts&TelecommunicationsXi’an 710061China School of Engineering and Computer Science Australian National UniversityCanberra 0200Australia

An image encryption algorithm is proposed in this paper based on a new four-dimensional hyperchaotic system,a neural mechanism,a Galois field and an improved Feistel block structure,which improves the efficiency and enhances the security of the encryption ***,a four-dimensional hyperchaotic system with a large key space and chaotic dynamics performance is proposed and combined with a cloud model,in which a more complex and random sequence is constructed as the key stream,and the problem of chaotic periodicity is ***,the key stream is combined with the neural mechanism,Galois field and improved Feistel block structure to scramble and diffuse the image ***,the experimental results and security analysis show that the encryption algorithm has a good encryption effect and high encryption efficiency,is secure,and can meet the requirements of practical applications.

关键词： image encryption neural mechanism hyperchaotic system

来源：评论

学校读者我要写书评

暂无评论

An Insight Into Communication Technologies Utilized in Smart Metering Systems 45

An Insight Into Communication Technologies Utilized in Smart...

引用

45th Jubilee International Convention on Information, Communication and Electronic Technology, MIPRO 2022

作者： Bacnar, David Leytner, Lolita Stojkovic, Nino Lerga, Jonatan University of Rijeka Faculty of Engineering Department of Computer Engineering Rijeka Croatia University of Rijeka Faculty of Engineering Department of Automation and Electronics Rijeka Croatia University of Angers Polytech Angers Automation and Computer Engineering Department Angers France

ISBN: (纸本)9789532331035

This paper presents an insight into the smart metering field. The paper address specifics of different network hierarchy structures (wide area networks, neighborhood area networks, and home area networks) and topologies used for this purpose (mesh, point-to-point, virtual point-to-point, or broadcast networks). Next, we focus on listing and describing some of the most commonly used communication technologies (both wired and wireless, such as power line communication, public switched telephone network, digital subscriber line, radiofrequency, in-home radio frequency, wireless local area network, or mobile network communications), their characteristics, advantages, and disadvantages utilized for remote and smart metering. © 2022 Croatian Society MIPRO.

关键词： Wide area networks

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：