检索结果-南通市图书馆

Synchronized perturbation elimination and DOA estimation via signal selection mechanism and parallel deep capsule networks in multipath environment

引用

Chinese Journal of Aeronautics 2021年第12期34卷 158-170页

作者： Ying CHEN Cong WANG Kunlai XIONG Zhitao HUANG State Key Laboratory of Complex Electromagnetic Environment Effects on Electronics and Information System National University of Defense TechnologyChangsha 410073China Institute of Information Fusion Naval Aeronautical and Astronautical UniversityYantai 264001China

State-of-the-art model-driven Direction-Of-Arrival(DOA)estimation methods for multipath signals face great challenges in practical application because of the dependence on the precise multipath *** this paper,we introduce a framework,based on deep learning,for synchronizing perturbation auto-elimination with effective DOA estimation in multipath ***,a signal selection mechanism is introduced to roughly locate specific signals to spatial subregion via frequency domain filters and compressive sensing-based ***,we set the mean of the correlation matrix’s row vectors as the input feature to construct the spatial spectrum by the corresponding single network within the parallel deep capsule *** proposed method enhances the generalization capability to untrained scenarios and the adaptability to non-ideal conditions,e.g.,lower SNRs,smaller snapshots,unknown reflection coefficients and perturbational steering vectors,which make up for the defects of the previous model-driven *** are carried out to demonstrate the superiority of the proposed method.

关键词： Deep capsule network Direction-Of-Arrival(DOA)estimation Multipath propagation parallel training Perturbation elimination

来源：

维普期刊数据库

同方期刊数据库评论

在线全文

学校读者我要写书评

暂无评论

A Reinforcement Learning training Acceleration Method Based on Knowledge Distillation

A Reinforcement Learning Training Acceleration Method Based ...

引用

第43届中国控制会议

作者： Zhiwen Wang Rujian Zhang Yanke Xu Qing Fei School of Automation Beijing Institute of Technology

The low efficiency of reinforcement learning training has long been a key constraint limiting its practical application in industrial domains. To address this issue, this paper proposes a parallel training framework named MCIC-GA3C. This framework employs knowledge distillation techniques to compress the knowledge of complex target policy models into smallsized behavioral policy networks in real-time. By accelerating the policy prediction process through model compression, the training efficiency of the algorithm is enhanced. However, the inconsistency between the behavioral policy and the target policy may result in policy deviation. To mitigate this issue, MCIC-GA3C introduces importance sampling methods. Simulated experiments are conducted in both Atari games and autonomously designed real-time drone path planning scenarios. Results demonstrate that using compressed behavioral policies for training increases round sampling speed by 17.95%. Furthermore,importance weight correction methods can partially address policy deviation issues, further enhancing the training efficiency of the algorithm. The proposed method also exhibits good training performance in autonomously developed simulation scenarios,validating its effectiveness and applicability.

关键词： Reinforcement Learning parallel training Knowledge Distillation Model Compression Path Planning

来源： cnki会议评论

在线全文

cnki会议

学校读者我要写书评

暂无评论

一种基于多支持向量机的并行增量学习方法(英文)

一种基于多支持向量机的并行增量学习方法(英文)

引用

第二十二届中国数据库学术会议

作者：张健沛李忠伟杨静哈尔滨工程大学计算机科学与技术学院

Support Vector Machine(SVM) can be trained and to learn incrementally when dealing with large-scale problems with batch training model,but the disadvantage is that its computation and storage requirements increase rapidly with the number of training *** overcome it,parallel training methods have been proposed by splitting the problem into smaller subsets and training a network.A parallel training algorithm with multiple SVM classifiers to incremental learning is proposed in this paper,in which multiple SVM classifiers are applied to batch model. Each classifier is trained by updating support vectors set with that of other classifiers to avoid the problem that the training results are subject to numbers of batches and state of data *** experiment results on realworld text dataset show that the parallel training algorithm with multiple SVM classifiers has more satisfying accuracy compared with batch training algorithm with a single SVM.

关键词： Support vector machine Incremental learning Multiple classifiers parallel training

来源： cnki会议评论

在线全文

cnki会议

学校读者我要写书评

暂无评论

欢迎您,

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

在线全文

在线全文

在线全文

请选择保存的检索档案：

请选择收藏分类：

通借通还

欢迎您,

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

在线全文

在线全文

在线全文

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：