检索结果-内蒙古大学图书馆

Adversarial Perturbation Defense on Deep Neural Networks

ACM COMPUTING SURVEYS 2021年第8期54卷 159-159页

作者： Zhang, Xingwei Zheng, Xiaolong Mao, Wenji Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China

Deep neural networks (DNNs) have been verified to be easily attacked by well-designed adversarial perturbations. Image objects with small perturbations that are imperceptible to human eyes can induce DNN-based image class classifiers towards making erroneous predictions with high probability. Adversarial perturbations can also fool real-world machine learning systems and transfer between different architectures and datasets. Recently, defense methods against adversarial perturbations have become a hot topic and attracted much attention. A large number of works have been put forward to defend against adversarial perturbations, enhancing DNN robustness against potential attacks, or interpreting the origin of adversarial perturbations. In this article, we provide a comprehensive survey on classical and state-of-the-art defense methods by illuminating their main concepts, in-depth algorithms, and fundamental hypotheses regarding the origin of adversarial perturbations. In addition, we further discuss potential directions of this domain for future researchers.

关键词： Adversarial perturbation defense deep neural networks security origin

来源：评论

学校读者我要写书评

暂无评论

AdvDiffuser: Generating Adversarial Safety-Critical Driving Scenarios via Guided Diffusion

arXiv

引用

arXiv 2024年

作者： Xie, Yuting Guo, Xianda Wang, Cong Liu, Kunhua Chen, Long School of Computer Science and Engineering Sun Yat-Sen University China School of Computer Science Wuhan University China State Key Laboratory for Management and Control of Complex Systems The Institute of Automation Chinese Academy of Sciences China School of Mechanical and Automotive Engineering Qingdao University of Technology China

Safety-critical scenarios are infrequent in natural driving environments but hold significant importance for the training and testing of autonomous driving systems. The prevailing approach involves generating safety-critical scenarios automatically in simulation by introducing adversarial adjustments to natural environments. These adjustments are often tailored to specific tested systems, thereby disregarding their transferability across different systems. In this paper, we propose AdvDiffuser, an adversarial framework for generating safety-critical driving scenarios through guided diffusion. By incorporating a diffusion model to capture plausible collective behaviors of background vehicles and a lightweight guide model to effectively handle adversarial scenarios, AdvDiffuser facilitates transferability. Experimental results on the nuScenes dataset demonstrate that AdvDiffuser, trained on offline driving logs, can be applied to various tested systems with minimal warm-up episode data and outperform other existing methods in terms of realism, diversity, and adversarial performance. © 2024, CC BY.

关键词： Automatic guided vehicles

来源：评论

学校读者我要写书评

暂无评论

Discretizing Continuous Action Space with Unimodal Probability Distributions for On-Policy Reinforcement Learning

arXiv

引用

arXiv 2024年

作者： Zhu, Yuanyang Wang, Zhi Zhu, Yuanheng Chen, Chunlin Zhao, Dongbin The Department of Control Science and Intelligent Engineering School of Management and Engineering Nanjing University Nanjing210093 China The Research Center for Novel Technology of Intelligent Equipment Nanjing University Nanjing210093 China The State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences Beijing100190 China The School of Artificial Intelligence University of Chinese Academy of Sciences Beijing100049 China

For on-policy reinforcement learning, discretizing action space for continuous control can easily express multiple modes and is straightforward to optimize. However, without considering the inherent ordering between the discrete atomic actions, the explosion in the number of discrete actions can possess undesired properties and induce a higher variance for the policy gradient estimator. In this paper, we introduce a straightforward architecture that addresses this issue by constraining the discrete policy to be unimodal using Poisson probability distributions. This unimodal architecture can better leverage the continuity in the underlying continuous action space using explicit unimodal probability distributions. We conduct extensive experiments to show that the discrete policy with the unimodal probability distribution provides significantly faster convergence and higher performance for on-policy reinforcement learning algorithms in challenging control tasks, especially in highly complex tasks such as Humanoid. We provide theoretical analysis on the variance of the policy gradient estimator, which suggests that our attentively designed unimodal discrete policy can retain a lower variance and yield a stable learning process. © 2024, CC BY.

关键词： Contrastive Learning

来源：评论

学校读者我要写书评

暂无评论

Deep learning radiomics of ultrasonography can predict response to neoadjuvant chemotherapy in breast cancer at an early stage of treatment: a prospective study

引用

EUROPEAN RADIOLOGY 2022年第3期32卷 2099-2109页

作者： Gu, Jionghui Tong, Tong He, Chang Xu, Min Yang, Xin Tian, Jie Jiang, Tianan Wang, Kun Zhejiang Univ Coll Med Affiliated Hosp 1 Dept Ultrasound Hangzhou 310003 Zhejiang Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst CAS Key Lab Mol Imaging Beijing 100190 Peoples R China Univ Chinese Acad Sci Sch Artificial Intelligence Beijing 100049 Peoples R China Beihang Univ Sch Med & Engn Beijing Adv Innovat Ctr Big Data Based Precis Med Beijing 100191 Peoples R China Zhejiang Prov Key Lab Pulsed Elect Field Technol Hangzhou Zhejiang Peoples R China

Objectives Breast cancer (BC) is the most common cancer in women worldwide, and neoadjuvant chemotherapy (NAC) is considered the standard of treatment for most patients with BC. However, response rates to NAC vary among patients, which leads to delays in appropriate treatment and affects the prognosis for patients who ineffectively respond to NAC. This study aimed to investigate the feasibility of deep learning radiomics (DLR) in the prediction of NAC response at an early stage. Methods In total, 168 patients with clinicopathologically confirmed BC were enrolled in this prospective study, from March 2016 to December 2020. All patients completed NAC treatment and underwent ultrasonography (US) at three time points (before NAC, after the second course, and after the fourth course). We developed two DLR models, DLR-2 and DLR-4, for predicting responses after the second and fourth courses of NAC. Furthermore, a novel deep learning radiomics pipeline (DLRP) was proposed for stepwise prediction of response at different time points of NAC administration. Results In the validation cohort, DLR-2 achieved an AUC of 0.812 (95% CI: 0.770-0.851) with an NPV of 83.3% (95% CI: 76.5-89.6). DLR-4 achieved an AUC of 0.937 (95% CI: 0.913-0.955) with a specificity of 90.5% (95% CI: 86.3-94.2). Moreover, 19 of 21 non-response patients were successfully identified by DLRP, suggesting that they could benefit from treatment strategy adjustment at an early stage of NAC. Conclusions The proposed DLRP strategy holds promise for effectively predicting NAC response at its early stage for BC patients. key Points We proposed two novel deep learning radiomics (DLR) models to predict response to neoadjuvant chemotherapy (NAC) in breast cancer (BC) patients based on US images at different NAC time points. Combining two DLR models, a deep learning radiomics pipeline (DLRP) was proposed for stepwise prediction of response to NAC. The DLRP may provide BC patients and physicians with an effective and fe

关键词： Breast cancer Deep learning Neoadjuvant chemotherapy Ultrasonography Treatment outcome

来源：评论

学校读者我要写书评

暂无评论

Path Following and Collision Avoidance of a Ribbon-Fin Propelled Underwater Biomimetic Vehicle-Manipulator System

SSRN

引用

SSRN 2022年

作者： He, Yanbing Dong, Xiang Wang, Yu Wang, Shuo Cheng, Long School of Electrical Engineering and Automation Anhui University Hefei230601 China State Key Laboratory of Management and Control for Complex Systems Institute of Automation Chinese Academy of Sciences China

This paper addresses the problem of path following and dynamic obstacle avoidance for underwater biomimetic vehicle-manipulator system (UBVMS). Firstly, the general kinematics and dynamics models of underwater vehicles are presented;then a nonlinear model predictive control (NMPC) scheme is employed to track a reference path and collision avoidance simultaneously. Moreover, to minimize the tracking error and a higher degree of robustness, improved extended state observers are used to estimate model uncertainties and disturbances to be fed into the NMPC prediction model. On top of this, the proposed method also considers the uncertainty of the state estimator, while combining a priori estimation of the Kalman filter to reasonably predict the position of dynamic obstacles during short periods. Finally, simulations and experimental results are carried out to assess the validity of the proposed method in case of disturbances and constraints. © 2022, The Authors. All rights reserved.

关键词： Vehicles

来源：评论

学校读者我要写书评

暂无评论

Switching Event-Triggered Adaptive Resilient Dynamic Surface control for Stochastic Nonlinear CPSs With Unknown Deception Attacks

引用

IEEE TRANSACTIONS ON CYBERNETICS 2023年第10期53卷 6562-6570页

作者： Niu, Ben Chen, Wendi Su, Wei Wang, Huanqing Wang, Ding Zhao, Xudong Shandong Normal Univ Sch Informat Sci & Engn Jinan 250014 Shandong Peoples R China Tianjin Univ Sch Elect Automat & Informat Engn Tianjin 300072 Peoples R China Bohai Univ Sch Math Sci Jinzhou 121000 Peoples R China Beijing Univ Technol Fac Informat Technol Beijing 100124 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China Dalian Univ Technol Fac Elect Informat & Elect Engn Dalian 116023 Peoples R China

This work concentrates on the adaptive resilient dynamic surface controller design problem for uncertain nonlinear lower triangular stochastic cyber-physical systems (CPSs) subject to unknown deception attacks based on a switching threshold event-triggered mechanism. The adverse effect of deception attacks on the stochastic CPSs is that the exact system state variables become unavailable. Furthermore, it should be emphasized that the coexistence of unknown nonlinearities, stochastic perturbations, and unknown sensor and actuator attacks makes it a very difficult and challenging event to implement the control design. To get the desired controller, radial basis function (RBF) neural networks (NNs) are introduced so that the design obstacle caused from the unknown nonlinearities can be easily solved. On this basis, in order to save resources and effectively transmit, the event-triggered control scheme based on a switching threshold strategy is further considered. In the backstepping design process, the dynamic surface control (DSC) technique is presented to deal with the issue of "explosion of complexity." By skillfully designing a new coordinate transformation and the attack compensators, the problem of unknown deception attacks is successfully handled. Under our proposed control scheme, all the closed-loop signals are bounded in probability and the stabilization errors converge to an adjustable neighborhood of the origin in probability. Finally, the simulation results on the double chemical reactor show the validity of the proposed design scheme.

关键词： Switches Stochastic processes Security Heuristic algorithms Actuators Nonlinear systems control design Dynamic surface control (DSC) stochastic nonlinear cyber-physical systems (CPSs) switching threshold event-triggered control unknown deception attacks

来源：评论

学校读者我要写书评

暂无评论

Decentralized robust optimal control for modular robot manipulators via critic-identifier structure-based adaptive dynamic programming

引用

NEURAL COMPUTING & APPLICATIONS 2020年第8期32卷 3441-3458页

作者： Dong, Bo Zhou, Fan Liu, Keping Li, Yuanchun Changchun Univ Technol Dept Control Sci & Engn Changchun 130012 Peoples R China Chinese Acad Sci Inst Automat State Key Lab Management & Control Complex Syst Beijing 100190 Peoples R China

This paper presents a decentralized robust optimal control method for modular robot manipulators (MRMs) via a critic-identifier structure-based adaptive dynamic programming (ADP) scheme. The robust control problem of MRMs is transformed into an optimal compensation control issue, which consists of model-based compensation control, identifier-based learning control and ADP-based optimal control. The dynamic model of MRMs is deployed for each joint module where the local dynamic information is utilized to design the model compensation controller. A neural network (NN) identifier is established to approximate the interconnected dynamic coupling. Based on ADP and local online policy iteration algorithm, the Hamiltonian-Jacobi-Bellman equation is solved by constructing a critic NN, and then the approximate optimal control policy derivation is possible. The closed-loop robotic system is asymptotic stable by the implementation of a set of developed decentralized control policies. Simulations are presented to demonstrate the effectiveness of the proposed method.

关键词： Modular robot manipulator Decentralized control Adaptive dynamic programming Optimal control Neural network Interconnected dynamic coupling

来源：评论

学校读者我要写书评

暂无评论

Research Progress of Parallel control and management

引用

IEEE/CAA Journal of Automatica Sinica 2020年第2期7卷 355-367页

作者： Gang Xiong Xisong Dong Hao Lu Dayong Shen State Key Laboratory of Management and Control for Complex Systems Institute of AutomationChinese Academy of SciencesBeijing 100190China School of Computer Science and Technology Beijing Institute of TechnologyBeijing 100081China College of Systems Engineering National University of Defense TechnologyChangsha 410073China

Based on ACP(artificial systems,computational experiments,and parallel execution)methodology,parallel control and management has become a popularly systematic and complete solution for the control and management of complex *** paper focuses on summarizing comprehensive review of the research literature of parallel control and management achieved in the recent years including the theoretical framework,core technologies,and the application *** future research,application directions,and suggestions are also discussed.

关键词： ACP methodology artificial systems computational experiments parallel control parallel management parallel systems

来源：评论

学校读者我要写书评

暂无评论

Geometry Problem Solving Based on Counter-Factual Evolutionary Reasoning 19

Geometry Problem Solving Based on Counter-Factual Evolutiona...

引用

19th IEEE International Conference on Automation Science and Engineering, CASE 2023

作者： Song, Bing Xiong, Gang Shen, Zhen Zhu, Fenghua Lv, Yisheng Ye, Peijun Institute of Automation Chinese Academy of Sciences The State Key Laboratory of Management and Control for Complex Systems Beijing China School of Artificial Intelligence University of Chinese Academy of Sciences Beijing100049 China Institute of Automation Chinese Academy of Sciences The Beijing Engineering Research Center of Intelligent Systems and Technology China Chinese Academy of Sciences The Guangdong Engineering Research Center of 3D Printing and Intelligent Manufacturing The Cloud Computing Center Dongguan523808 China Qingdao Academy of Intelligent Industries Intelligent Manufacturing Center Qingdao266000 China

ISBN: (纸本)9798350320695

As a representative topic in natural language processing and automated theorem proving, geometry problem solving requires an abstract problem understanding and symbolic reasoning. A major challenge here is to find a feasible reasoning sequence that is consistent with given axioms and the theorems already proved. Most recent methods have exploited neural network-based techniques to automatically discover eligible solving steps. Such a kind of methods, however, is greatly impacted by the expert solutions for training. To improve the accuracy, this paper proposes a new method called counterfactual evolutionary reasoning, which uses a generative adversarial network to generate initial reasoning sequences and then introduces counterfactual reasoning to explore potential solutions. By directly exploring theorem candidates rather than the neural network selection, the new method can sufficiently extend the searching space to get a more appropriate reasoning step. Through comparative experiments on the recent proposed Geometry3k, the largest geometry problem solving dataset, our method generally achieves a higher accuracy than most previous methods, bringing an overall improvement about 4.4% compared with the transformer models. © 2023 IEEE.

关键词： Geometry

来源：评论

学校读者我要写书评

暂无评论

Multi-Task Learning-Enabled Automatic Vessel Draft Reading for Intelligent Maritime Surveillance

arXiv

引用

arXiv 2023年

作者： Qu, Jingxiang Liu, Ryan Wen Zhao, Chenjie Guo, Yu Xu, Sendren Sheng-Dong Zhu, Fenghua Lv, Yisheng The Hubei Key Laboratory of Inland Shipping Technology School of Navigation Wuhan University of Technology Wuhan430063 China The State Key Laboratory of Maritime Technology and Safety Wuhan430063 China The Automation and Control Center The Graduate Institute of Automation and Control National Taiwan University of Science and Technology Taipei106335 Taiwan The State Key Laboratory for Management and Control of Complex Systems Institute of Automation Chinese Academy of Sciences Beijing100190 China

The accurate and efficient vessel draft reading (VDR) is an important component of intelligent maritime surveillance, which could be exploited to assist in judging whether the vessel is normally loaded or overloaded. The computer vision technique with an excellent price-to-performance ratio has become a popular medium to estimate vessel draft depth. However, the traditional estimation methods easily suffer from several limitations, such as sensitivity to low-quality images, high computational cost, etc. In this work, we propose a multi-task learning-enabled computational method (termed MTL-VDR) for generating highly reliable VDR. In particular, our MTL-VDR mainly consists of four components, i.e., draft mark detection, draft scale recognition, vessel/water segmentation, and final draft depth estimation. We first construct a benchmark dataset related to draft mark detection and employ a powerful and efficient convolutional neural network to accurately perform the detection task. The multi-task learning method is then proposed for simultaneous draft scale recognition and vessel/water segmentation. To obtain more robust VDR under complex conditions (e.g., damaged and stained scales, etc.), the accurate draft scales are generated by an automatic correction method, which is presented based on the spatial distribution rules of draft scales. Finally, an adaptive computational method is exploited to yield an accurate and robust draft depth. Extensive experiments have been implemented on the realistic dataset to compare our MTL-VDR with state-of-the-art methods. The results have demonstrated its superior performance in terms of accuracy, robustness, and efficiency. The computational speed exceeds 40 FPS, which satisfies the requirements of real-time maritime surveillance to guarantee vessel traffic safety. Copyright © 2023, The Authors. All rights reserved.

关键词： Image segmentation

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：