检索结果-内蒙古大学图书馆

11th International Conference on Image and Graphics, ICIG 2021

作者： Zhang, Haitao Shi, Yucheng Dong, Benyu Han, Yahong Li, Yuanzhang Kuang, Xiaohui College of Intelligence and Computing and Tianjin Key Lab of Machine Learning Tianjin University Tianjin China Beijing Institute of Technology Beijing China National Key Laboratory of Science and Technology on Information System Security Beijing China

ISBN: (纸本)9783030873578

Due to the existence of adversarial attacks, various applications that employ deep neural networks (DNNs) have been under threat. Adversarial training enhances robustness of DNN-based systems by augmenting training data with adversarial samples. Projected gradient descent adversarial training (PGD AT), one of the promising defense methods, can resist strong attacks. We propose "free" adversarial training with layerwise heuristic learning (LHFAT) to remedy these problems. To reduce heavy computation cost, we couple model parameter updating with projected gradient descent (PGD) adversarial example updating while retraining the same mini-batch of data, where we "free" and unburden extra updates. Learning rate reflects weight updating speed. Weight gradient indicates weight updating efficiency. If weights are frequently updated towards opposite directions in one training epoch, then there are redundant updates. For higher level of weight updating efficiency, we design a new learning scheme, layerwise heuristic learning, which accelerates training convergence by restraining redundant weight updating and boosting efficient weight updating of layers according to weight gradient information. We demonstrate that LHFAT yields better defense performance on CIFAR-10 with approximately 8% GPU training time of PGD AT and LHFAT is also validated on ImageNet. We have released the code for our proposed method LHFAT at https://***/anonymous530/LHFAT. © 2021, Springer Nature Switzerland AG.

关键词： Deep neural networks

来源：评论

学校读者我要写书评

暂无评论

Accomplishing robot grasping task rapidly via adversarial training

Accomplishing robot grasping task rapidly via adversarial tr...

引用

2019 IEEE International Conference on Real-Time computing and Robotics, RCAR 2019

作者： Zuo, Guoyu Lu, Jiahao Chen, Kexin Yu, Jianjun Huang, Xiangsheng Faculty of Information Technology Beijing University of Technology Beijing100124 China Beijing Key Laboratory of Computing Intelligence and Intelligent Systems Beijing100124 China Institute of Automation Chinese Academy of Sciences Beijing100190 China

ISBN: (纸本)9781728137261

This paper proposes a robotic imitation learning method which integrates the deterministic off-policy reinforcement learning and generative adversarial network. This method allows the robot to implement the grasping task rapidly by learning the reward function from the demonstration data. Firstly, the discriminator is used to learn the reward function from demonstrations, which can guide the generator to complete the robot grasping task. Secondly, the deep deterministic policy gradient method is used as the generator for learning action policy on the basis of discriminator. In particular, the demonstration data is also input into the generator to ensure its performance. Finally, three experiments on the Push and Pick- and-Place tasks are conducted in the GYM robotic environment. Results show that the learning speed of our method is much faster than the stochastic GAIL method, and it can effectively train from the demonstration data in different states of the task. The proposed method can complete the robot grasping task without environmental reward quickly and improve the stability of the training process. © 2018 IEEE

关键词： Demonstrations

来源：评论

学校读者我要写书评

暂无评论

Dynamics Modeling with Realistic Constraints for Trajectory Tracking Control of Manipulator∗

Dynamics Modeling with Realistic Constraints for Trajectory ...

引用

2022 IEEE International Conference on Real-Time computing and Robotics, RCAR 2022

作者： Liu, Lu Zuo, Guoyu Li, Jiangeng Li, Jianfeng Beijing Key Laboratory of Computing Intelligence and Intelligent Systems Beijing100124 China Beijing University of Technology Faculty of Materials and Manufacturing Beijing100124 China Faculty of Materials and Manufacturing Beijing University of Technology Beijing100124 China

ISBN: (数字)9781665469838

ISBN: (纸本)9781665469838

To meet the practical operation requirements of manipulators in industry, service and collaboration scenarios, trajectory tracking accuracy and system stability of manipulators are usually regarded as critical control objectives. This paper proposes a novel manipulator control method with three-level constraints, called the real deep Lagrangian network (Real-DeLaN). In this method, real data (also called physical data) are collected from the physical manipulator and utilized to train the Lagrangian dynamics network model to improve the migration ability from simulation to reality. The torque output of the dynamic model is corrected by the real-Time calculation of friction in the network. The contact force on the end effector of the manipulator is compensated based on the principle of virtual displacement work. The experimental results show that Real-DeLaN can better control the joints to perform trajectory tracking, reduce the friction error of the manipulator, and show better anti-interference ability. © 2022 IEEE.

关键词： Manipulators

来源：评论

学校读者我要写书评

暂无评论

An Efficient Lightweight Spatio-temporal Attention Module for Action Recognition 11

An Efficient Lightweight Spatio-temporal Attention Module fo...

引用

11th International Conference on computing and Pattern Recognition, ICCPR 2022

作者： Sun, Zhonghua Dai, Meng Yi, Ziwen Wang, Tianyi Feng, Jinchao Jia, Kebin Faculty of Information Technology Beijing University of Technology Beijing Laboratory of Advanced Information Networks Beijing China Faculty of Information Technology Beijing University of Technology China Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing University of Technology China

ISBN: (纸本)9781450397056

Effective feature learning is one of the prime components for human action recognition algorithm. Three-dimensional convolutional neural network (3D CNN) can directly extract spatio-temporal features, however it is insufficient to capture the most discriminative part of the action video. The redundant spatial regions within and between temporal frames would weak the descriptive ability of the 3D CNN model. To address this problem, we propose a lightweight spatio-temporal attention module (ST-AM), composed of spatial attention module (SAM) and temporal attention module (TAM). SAM and TAM can effectively encode the semantic spatial areas and suppress the redundant temporal frames to reduce misclassification. The proposed SAM and TAM have complementary effects and can be easily embedded into the existing 3D CNN action recognition model. Experiment on UCF-101 and HMDB-51 datasets shows that the ST-AM embedded model achieves impressive performance on action recognition task. © 2022 ACM.

关键词： Semantics

来源：评论

学校读者我要写书评

暂无评论

A fast growing cascade neural network for BOD estimation 34

A fast growing cascade neural network for BOD estimation

引用

34th Chinese Control Conference, CCC 2015

作者： Fanjun, Li Junfei, Qiao Wei, Zhang College of Electronic Information and Control Engineering Beijing University of Technology Beijing China School of Mathematical Science University of Jinan Shandong China Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing China

ISBN: (纸本)9789881563897

In this paper, a fast growing cascade neural network (FGCNN) is proposed, as a software sensor, to rapidly estimate the biochemical oxygen demand (BOD) in wastewater treatment plants (WWTPs). Firstly, a novel method, based on the orthogonal least squares (OLS), is put forward to add input and hidden units to the existing network one by one. Every unit added to the network affords the maximal reduction of the sum of squared errors (SSE). Then, the FGCNN incrementally updates its output weights by iterations without gradients and generalized inverses, while the other weights remain unchanged during the growth of the network. The simple and effective training method make the FGCNN learn extremely fast. Finally, the proposed FGCNN is applied to estimate the BOD in WWTPs using other easy-to-measure or secondary variables. The experiment results show that the FGCNN has better performance on real-time estimation of BOD than other similar methods. © 2015 Technical Committee on Control Theory, Chinese Association of Automation.

关键词： Artificial neural network biochemical oxygen demand software sensor wastewater treatment plant

来源：评论

学校读者我要写书评

暂无评论

A local search strategy based on density for NSGA2 algorithm

引用

Kongzhi yu Juece/Control and Decision 2018年第1期33卷 60-66页

作者： Li, San-Yi Li, Wen-Jing Qiao, Jun-Fei College of Automation Beijing University of Technology Beijing100124 China Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing100124 China

In order to reduce the amount of calculation and keep the advantage of local search strategy simultaneously, this paper proposes a kind of local search method based on density for the NSGA2(NSGA2-DLS) algorithm. Firstly, the sparse degree of each solution is evaluated with the density of each solution in the solution space. Then the non-dominated solution with the smallest sparse degree is defined as the sparse solution, and the sparse solutionig is searched around locally during every genetic process. The NSGA2-DLS algorithm adopts extreme optimization strategy and random search strategy simultaneously to improve the quality of solutions and convergence rate. The performance of the NSGA2-DLS algorithm is compared with the performance of the baseline NSGA2 algorithm and two other reported local search NSGA2 algorithms for multi-objective test problems, including ZDT and DTLZ functions. The simulation results show that the solution quality and the calculation amount of the NSGA2-DLS algorithm are much better than that of other methods. © 2018, Editorial Office of Control and Decision. All right reserved.

关键词： Multiobjective optimization

来源：评论

学校读者我要写书评

暂无评论

Perceptual Quality Evaluation of Corrupted Industrial Images 18th

Perceptual Quality Evaluation of Corrupted Industrial Images

引用

18th International Forum of Digital Multimedia Communication, IFTC 2021

作者： Gong, Yafei Peng, Chenchen Liu, Jing Zhou, Chengxu Liu, Hongyan Faculty of Information Technology Beijing University of Technology Beijing China Engineering Research Center of Intelligent Perception and Autonomous Control Ministry of Education Beijing China Beijing Laboratory of Smart Environmental Protection Beijing China Beijing Key Laboratory of Computational Intelligence and Intelligent System Beijing China Beijing Artificial Intelligence Institute Beijing China Faculty of Science Beijing University of Technology Beijing China

ISBN: (纸本)9789811922657

In recent years, computer vision applied in industrial application scenarios has been attracting attention. A lot of work has been made to present approaches based on visual perception, ensuring the safety during the processes of industrial production. However, much less effort has been done to assess the perceptual quality of corrupted industrial images. In this paper, we construct an Industrial Scene Image Database (ISID), which contains 3000 distorted images generated through applying different levels of distortion types to each of the 50 source images. Then, the subjective experiment is carried out to gather the subjective scores in a well-controlled laboratory environment. Finally, we perform comparison experiments on ISID database to investigate the performance of some objective image quality assessment algorithms. The experimental results show that the state-of-the-art image quality assessment methods have difficulty in predicting the quality of images that contain multiple distortion types. © 2022, Springer Nature Singapore Pte Ltd.

关键词： Database systems

来源：评论

学校读者我要写书评

暂无评论

Combine Neural Radiance Fields in Different Reference Coordinates

Combine Neural Radiance Fields in Different Reference Coordi...

引用

2023 China Automation Congress, CAC 2023

作者： Zhang, Weili Li, Zhongheng Guo, Yu Wang, Fei Institute of Artificial Intelligence and Robotics Xi'An Jiaotong University Xi'an China Beijing Institute of Spacecraft System Engineering Beijing Key Laboratory of Intelligent Space Robotic Systems Technology and Applications Beijing China

ISBN: (纸本)9798350303759

Neural Radiance Fields (NeRF) leverage deep learning networks to restore high-quality implicit representations of 3D scenes from 2D images, surpassing traditional texture mapping in rendering quality. Typically, reconstructed images by NeRF exist in a reference coordinate system, necessitating consideration of each surface's information in a single shot. Overlapping parts in adjacent images are addressed by obtaining the camera pose through sparse COLMAP reconstruction. However, occluded regions, caused by contact with support surfaces, pose challenges for complete reconstruction. To address this, we propose a scene fusion algorithm based on NeRF, facilitating the amalgamation of neural radiance fields under different reference coordinate systems. This reduces data requirements and shooting conditions, enhancing the implicit 3D reconstruction of real objects. By estimating the translation and rotation relationship between image capture angles and the NeRF model, we fuse neural radiance fields in distinct reference coordinate systems, completing object surface information. Evaluation is conducted on both NeRF synthetic dataset and real objects. © 2023 IEEE.

关键词： Cameras

来源：评论

学校读者我要写书评

暂无评论

Feature Subset Discernibility Evaluation Method for Upper Limb Rehabilitation Training Based on the Discernibility of Relative Distance

Feature Subset Discernibility Evaluation Method for Upper Li...

引用

IEEE International Conference on Mechatronics and Automation

作者： Guoyu Zuo Zhaokun Xu Jiahao Lu Beijing Key Laboratory of Computing Intelligence and Intelligent Systems Beijing China

In this paper, a feature subset discernibility hybrid evaluation method based on the discernibility of relative distance and support vector machine (DRD-SVM) is proposed for the feature selection problem of the upper limb rehabilitation training motion of Brunnstrom 4-5 stage patients, in which a relative distance is introduced into evaluating the discernibility between classes considering the joint effect of both candidate and selected features. First, a feature subset search strategy is used to search a set of candidate feature subsets. Then the DRD is used to evaluate the candidate feature subsets, the best subset is selected as a new selected feature subset, and the feature subset with the best performance of SVM classification is selected as the optimal feature subset. Finally, feature selection experiment was carried out on upper limb routine rehabilitation training samples of the Brunnstrom 4-5 stage. The experimental results shows that, compared with the F-score method and the DFS one, the proposed method can obtain the feature subsets with higher accuracy and smaller feature dimension, which improves its effectiveness and feasibility.

关键词： Training Feature extraction Support vector machines Filtering theory Filtering algorithms Sensors Search problems

来源：评论

学校读者我要写书评

暂无评论

Data protection and provenance in cloud of things environment: Research challenges

引用

International Journal of Information and Computer Security 2020年第4期12卷 416-435页

作者： Wang, Chundong Yang, Lei Guo, Hao Wan, Fujin Key Laboratory of Computer Vision and System Tianjin Key Laboratory of Intelligence Computing and Novel Software Technology Ministry of Education Tianjin University of Technology Tianjin China Global Energy Internet Research Institute Beijing University of Technology Tianjin China Department of College of Computer and Control Engineering Nankai University Tianjin China

Internet of things are increasingly being deployed over the cloud (also referred to as cloud of things) to provide a broader range of services. However, there are serious challenges of CoT in the data protection and security provenance. This paper proposes a data privacy protection and provenance model (DPSPM) based on CoT. It can protect the privacy data of the users and trace the source of leaked data. In detail, security encryption and watermarking algorithms are proposed. Meanwhile, we use the improved k-anonymity data masking algorithm and pseudo-row watermarking algorithm in this scheme. Those algorithms can carry out security control over the whole process of data publishing, especially in data encryption, data masking and provenance verification. Finally, the experimental results show that our scheme has good efficiency. It is proved that the data masking time is proportional to the parameters k and L, the results also show good robustness to the common database watermarking attacks. © 2020 Inderscience Enterprises Ltd.

关键词： Data privacy

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：