检索结果-内蒙古大学图书馆

arXiv 2023年

作者： Yang, Wenyan Wang, Huiling Cai, Dingding Pajarinen, Joni Kämäräinen, Joni-Kristian Computing Sciences Tampere University Finland Department of Electrical Engineering and Automation Aalto University Finland

Offline goal-conditioned reinforcement learning (GCRL) can be challenging due to overfitting to the given dataset. To generalize agent’s skills outside the given dataset, we propose a goal-swapping procedure that generates additional trajectories. To alleviate the problem of noise and extrapolation errors, we present a general offline reinforcement learning method called deterministic Q-advantage policy gradient (DQAPG). In the experiments, DQAPG outperforms state-of-the-art goal-conditioned offline RL methods in a wide range of benchmark tasks, and goal-swapping further improves the test results. It is noteworthy, that the proposed method obtains good performance on the challenging dexterous in-hand manipulation tasks for which the prior methods failed. © 2023, CC BY-NC-SA.

关键词： Reinforcement learning

来源：评论

学校读者我要写书评

暂无评论

PRIORITIZED OFFLINE GOAL-SWAPPING EXPERIENCE REPLAY

arXiv

引用

arXiv 2023年

作者： Yang, Wenyan Pajarinen, Joni Cai, Dingding Kämäräinen, Joni-Kristian Computing Sciences Tampere University Finland Department of Electrical Engineering and Automation Aalto University Finland

In goal-conditioned offline reinforcement learning, an agent learns from previously collected data to go to an arbitrary goal. Since the offline data only contains a finite number of trajectories a main challenge is how to generate more data. Goal-swapping generates additional data by switching trajectory goals but while doing so produces a large number of invalid trajectories. To address this issue we propose prioritized goal-swapping experience replay (PGSER). PGSER uses a pre-trained Q function to assign higher priority weights to goal-swapped transitions that allow reaching the goal. In experiments, PGSER significantly improves over baselines in a wide range of benchmark tasks. © 2023, CC BY-NC-SA.

关键词： Trajectories

来源：评论

学校读者我要写书评

暂无评论

HDCBAN: Hybrid Neural Network for Network Intrusion Detection System

HDCBAN: Hybrid Neural Network for Network Intrusion Detectio...

引用

International Conference on Computers, Communications, and Systems (ICCCS)

作者： Bo Xie Mingdi Xu Chaoyang Jin Feng Cui Zhengxiao Li Haipeng Fan Computing Environment Department Wuhan Institute of Digital Engineering Wuhan China School of Automation China University of Geosciences Wuhan China

ISBN: (数字)9798350350210

ISBN: (纸本)9798350350227

Intrusion Detection Systems (IDS) play a crucial role in safeguarding network security perimeters by monitoring network traffic and system behavior in real time. The advent of deep learning, particularly through the application of Convolutional Neural Networks (CNNs) and Long Short-Term Memory (LSTM) networks, has significantly enhanced IDS's ability to detect network threats by leveraging autonomous learning and generalization capabilities. However, deep learning-based IDSs encounter challenges, including the necessity for extensive data labeling and the computational resources required, which pose obstacles to their widespread adoption and effectiveness. Nonetheless, deep learning-based IDS encounter challenges, including the requirement for extensive labeled datasets and significant computational resources. This paper proposes a novel hybrid IDS framework — HDCBAN that integrates deep convolutional neural networks, bidirectional LSTM networks, and the AlexNet model to efficiently predict and classify malicious network activities. The HDCBAN model employs deep CNNs to capture local features through convolution, bidirectional LSTMs to seize temporal features for enhanced performance and prediction capabilities, and AlexNet to augment classification efficacy through feature extraction. The effectiveness of this hybrid approach was assessed using real-world datasets, including the publicly available CSE-CIC-IDS 2017, CSE-CIC-IDS 2018, and NSL-KDD datasets, with preprocessing to optimize model performance. Experimental results, validated through 10-fold cross-validation, reveal that our model achieves a malicious attack detection rate and accuracy of up to 99.88% for CSE-CIC-IDS and 99.93% for NSL-KDD, thereby outperforming existing models in detection rate, accuracy, and reducing false positives.

关键词： Training Accuracy Computational modeling Scalability Training data Telecommunication traffic Predictive models

来源：评论

学校读者我要写书评

暂无评论

Control Strategy based on Frequency Response for a Single-Phase Full- Bridge Series Active Filter 16

Control Strategy based on Frequency Response for a Single-Ph...

引用

16th Brazilian Power Electronics Conference, COBEP 2021

作者： Curi Busarello, Tiago Davi Automation and Computing Engineering Federal University of Santa Catarina Department of Control Blumenau Brazil

ISBN: (纸本)9781665421294

Series active filters are a mature technology and well covered in the literature. However, they are usually designed for three-phase systems employing control strategies in dq-rotating reference frame. As a result, proportional-integrator controllers are used. However, they are not suited when the system has abnormal conditions. Moreover, with the high penetration of distribution generators, single-phase distribution voltage is being deteriorated. This paper presents a control strategy based on frequency response for a single-phase full bridge series active filter. The series active filter is controlled in voltage mode through a digital lead-lag controller. The controller parameters are obtained by frequency response technique using the k-factor methodology and later discretized via Backward-Euler approximation. The series active filter is tested in a system with nonlinear loads, distortions at the distribution voltage and under disturbance at the DC-link. The proposal was verified under simulation results and under Hardware-in-Loop Typhoon HIL402 with external microcontroller TMS320F28335. Results showed the effectiveness of the proposed control strategy. © 2021 IEEE.

关键词： Active filters

来源：评论

学校读者我要写书评

暂无评论

Estimation of Cotton Boll Number and Main Stem Length Based on 3D Gaussian Splatting

Estimation of Cotton Boll Number and Main Stem Length Based ...

引用

2024 American Society of Agricultural and Biological Engineers Annual International Meeting, ASABE 2024

作者： Jiang, Lizhi Li, Changying Sun, Jin Chee, Peng Fu, Longsheng College of Mechanical and Electronic Engineering Northwest A&F University Shaanxi Yangling China Bio-Sensing Automation and Intelligence Laboratory Department of Agricultural and Biological Engineering University of Florida GainesvilleFL United States College of Computing University of Georgia AthensGA United States Department of Crop and Soil Science University of Georgia TiftonGA United States

ISBN: (纸本)9798331302214

Cotton is an important economic crop widely grown worldwide to produce textiles. Breeding programs aim to select genotypes with desirable architectural traits to develop new varieties with higher yields. To overcome time-consuming and labor-intensive manual measurements, this study developed a 3D Gaussian splatting (3DGS) based method to measure cotton phenotypic traits via instance segmentation. First, a 360-degree video of a single cotton plant was captured, and 2D images were extracted uniformly from the video. COLMAP was utilized to estimate the camera poses of the images and reconstruct the sparse point cloud. Second, the sparse point cloud was initialized as 3D Gaussians defined by position, covariance matrix, and opacity, and combined with camera poses to train the 3DGS model for the cotton plant scene. Next, to extract regions of interest from 2D images, the YOLOv8x instance segmentation model was used to generate 2D masks of the objects of interest. Then, segment any 3D Gaussians (SAGA) combines the generated initial 2D reference masks to obtain the 3DGS model of individual cotton bolls and the main stem. Finally, the phenotypic traits of the cotton plant, such as cotton boll number, and main stem length, were estimated. By comparing with the ground truth, the mean absolute percentage error (MAPE) for cotton boll counting and main stem length reached 11.43% and 10.45%, respectively. The methodology introduced in this study is a novel approach to generating new synthesized image perspectives and transforming 2D masks into 3D space. Accurate estimation of the phenotypic traits of cotton plants will benefit cotton breeding programs. © 2024 ASABE Annual International Meeting. All rights reserved.

关键词： Cotton

来源：评论

学校读者我要写书评

暂无评论

MAIDS - A Framework for the Development of Multi-Agent Intentional Dialogue Systems 22

MAIDS - A Framework for the Development of Multi-Agent Inten...

引用

22nd International Conference on Autonomous Agents and Multiagent Systems, AAMAS 2023

作者： Engelmann, Débora C. Panisson, Alison R. Vieira, Renata Hübner, Jomi Fred Mascardi, Viviana Bordini, Rafael H. School of Technology PUCRS Porto Alegre Brazil DIBRIS University of Genoa Genoa Italy Department of Computing UFSC Araranguá Brazil CIDEHUS University of Évora Évora Portugal Automation and Systems Engineering UFSC Florianópolis Brazil

This paper introduces a framework for programming highly sophisticated multi-agent dialogue systems. The framework is based on a multi-part agent belief base consisting of three components: (i) the main component is an extension of an agent-oriented programming belief base for representing defeasible knowledge and, in particular, argumentation schemes;(ii) an ontology component where existing OWL ontologies can be instantiated;and (iii) a theory of mind component where agents keep track of mental attitudes they ascribe to other agents. The paper formalises a structured argumentation-based dialogue game where agents can "digress" from the main dialogue into subdialogues to discuss ontological or theory of mind issues. We provide an example of a dialogue with an ontological digression involving humans and agents, including a chatbot that we developed to support bed allocation in a hospital. The example is used to show that our framework supports all features of recent desiderata for future dialogue systems. We also report an initial evaluation of the chatbot carried out by domain experts. © 2023 International Foundation for Autonomous Agents and Multiagent Systems (***). All rights reserved.

关键词： Ontology

来源：评论

学校读者我要写书评

暂无评论

Secure Data Sharing Method Based on Federated Learning in IoVs

Secure Data Sharing Method Based on Federated Learning in Io...

引用

IEEE International Conference on Communications (ICC)

作者： Shu-Jia Fan Jia-Wen Xue School of Computing and Information Technology University of Wollongong Australia Department of Automation & Information Engineering Xi'an University of Technology Xi'an China

ISBN: (数字)9781728190549

ISBN: (纸本)9781728190556

In the realm of the Internet of Vehicles (IoVs), a broad spectrum of information encompassing vehicle operation, traffic conditions, and incidents can be exchanged between vehicles and infrastructure. This data plays a pivotal role in enhancing decision-making processes and holds considerable importance for the advancement of autonomous driving and traffic management. Utilizing machine learning (ML) methods, vehicles can assimilate insights from road and environmental data, subsequently sharing trained model parameters. Addressing concerns related to data security and user privacy during the data-sharing process in IoVs, a framework called DBL is introduced. This framework leverages blockchain and federated learning to validate local model updates. It facilitates the updating and exchange of local ML models among multiple participants, with the option to upload these models to Road Side Units (RSUs). Subsequently, RSUs scrutinize the local ML models, identifying and discarding any malicious ones while aggregating the legitimate ones to generate the global ML model. Simulation results demonstrate that the proposed DBL framework not only safeguards user data privacy but also enhances model accuracy.

关键词： Training Data privacy Federated learning Data security Simulation Roads Road side unit

来源：评论

学校读者我要写书评

暂无评论

Monocular Visual-Inertial and Robotic-Arm Calibration in a Unifying Framework

引用

IEEE/CAA Journal of Automatica Sinica 2022年第1期9卷 146-159页

作者： Yinlong Zhang Wei Liang Mingze Yuan Hongsheng He Jindong Tan Zhibo Pang Key Laboratory of Networked Control Systems Shenyang Institute of AutomationChinese Academy of SciencesShenyang 110016 State Key Laboratory of Robotics Shenyang Institute of AutomationChinese Academy of SciencesShenyang 110016 Institutes for Robotics and Intelligent Manufacturing Chinese Academy of SciencesShenyang 110169China School of Computing Wichita State UniversityWichita 67260 USA Department of Mechanical Aerospace and Biomedical EngineeringUniversity of TennesseeTN 37996 USA Department of Automation Technology ABB Corporate Research SwedenVasteras 72178Sweden Department of Intelligent Systems Royal Institute of Technology(KTH)Stockholm 11428Sweden

Reliable and accurate calibration for camera,inertial measurement unit(IMU)and robot is a critical prerequisite for visual-inertial based robot pose estimation and surrounding environment ***,traditional calibrations suffer inaccuracy and *** address these problems,this paper proposes a monocular visual-inertial and robotic-arm calibration in a unifying *** our method,the spatial relationship is geometrically correlated between the sensing units and robotic *** decoupled estimations on rotation and translation could reduce the coupled errors during the ***,the robotic calibration moving trajectory has been designed in a spiral pattern that enables full excitations on 6 DOF motions repeatably and *** calibration has been evaluated on our developed *** the experiments,the calibration achieves the accuracy with rotation and translation RMSEs less than 0.7°and 0.01 m,*** comparisons with state-of-the-art results prove our calibration consistency,accuracy and effectiveness.

关键词： Calibration inertial measurement unit(IMU) monocular camera robotic arm spiral moving trajectory

来源：评论

学校读者我要写书评

暂无评论

A Fine-Grained Hemispheric Asymmetry Network for Accurate and Interpretable Eeg-Based Emotion Classification

SSRN

引用

SSRN 2024年

作者： Yan, Ruofan Lu, Na Yan, Yuxuan Niu, Xu Wu, Jibin Systems Engineering Institute School of Automation Science and Engineering Xian Jiaotong University China Department of Computing The Hong Kong Polytechnic University Hong Kong

Recent advancements in neuroscience have highlighted the significance of the asymmetricdiscrepancy between brain hemispheres, particularly within specific frequency bands, in capturingcrucial information for emotion recognition. Building upon this understanding, we introducethe Fine-grained Hemispheric Asymmetry Network (FG-HANet) designed to effectively extractdiscriminative hemispheric asymmetry features from EEG data for emotion recognition. In pursuitof capturing fine-grained spectral features, the FG-HANet incorporates Finite Impulse Response(FIR) filters at a granularity as precise as 2Hz, allowing for the initialization of spectral filterswithin the network. To optimize the learning of fine-grained hemispheric asymmetry features,we propose a three-stage training pipeline specifically tailored for the FG-HANet. Experimentalevaluations conducted on two datasets, namely SEED and SEED-IV, demonstrate the superiorclassification accuracies achieved by the FG-HANet, yielding 97.33% and 90.38% accuracy,respectively. Our results showcase the effectiveness of the proposed network in leveragingfine-grained hemispheric asymmetry features for EEG-based emotion recognition. Additionally,inspiring findings in neroscience have been found in this study. © 2024, The Authors. All rights reserved.

关键词： Emotion Recognition

来源：评论

学校读者我要写书评

暂无评论

Distributed Parameter Estimation under Byzantine Nodes in Sensor Networks

引用

IEEE Transactions on Network Science and engineering 2025年

作者： Yan, Jiaqi Li, Kuo Ishii, Hideaki Beihang University School of Automation Science and Electrical Engineering Beijing100191 China Tsinghua University Center for Intelligent and Networked System Department of Automation BNRist Beijing100084 China The University of Tokyo Department of Information Physics and Computing Tokyo113-8656 Japan

In this paper, we study the problem of parameter estimation in a sensor network, where the measurements and updates of some sensors might be arbitrarily manipulated by adversaries. Despite the presence of such misbehaviors, normally behaving sensors make successive observations of an unknown d-dimensional vector parameter and aim to infer its true value by cooperating with their neighbors over a directed communication graph. To this end, by leveraging the so-called dynamic regressor extension and mixing algorithm, we transform the problem of estimating the vector parameter to that of estimating d scalar ones. For each of the scalar problem, we propose a resilient combine-then-adapt (RCTA) diffusion algorithm, where each normal sensor performs a resilient combination to discard the suspicious estimates in its neighborhood and to fuse the remaining values, alongside an adaptation step to process its streaming observations. With a low computational cost, the proposed estimator guarantees that each normal sensor exponentially estimates the true parameter under some conditions. In particular, they consist of a sharp condition on the network topology and an excitation condition that requires only a subset of normal sensors to be sufficiently excited. It is further demonstrated that the proposed algorithm is robust to time delays and switching interactions. © 2025 IEEE.

关键词： Directed graphs

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：