检索结果-南通市图书馆

Low rank optimization for efficient deep learning:making a balance between compact architecture and fast training

Journal of Systems Engineering and Electronics 2024年第3期35卷 509-531,F0002页

作者： OU Xinwei CHEN Zhangxin ZHU Ce LIU Yipeng School of Information and Communication Engineering University of Electronic Science and Technology of ChinaChengdu 611731China

deep neural networks(DNNs)have achieved great success in many data processing ***,high computational complexity and storage cost make deep learning difficult to be used on resource-constrained devices,and it is not environmental-friendly with much power *** this paper,we focus on low-rank optimization for efficient deep learning *** the space domain,DNNs are compressed by low rank approximation of the network parameters,which directly reduces the storage requirement with a smaller number of network *** the time domain,the network parameters can be trained in a few subspaces,which enables efficient training for fast *** model compression in the spatial domain is summarized into three categories as pre-train,pre-set,and compression-aware methods,*** a series of integrable techniques discussed,such as sparse pruning,quantization,and entropy coding,we can ensemble them in an integration framework with lower computational complexity and *** addition to summary of recent technical advances,we have two findings for motivating future *** is that the effective rank,derived from the Shannon entropy of the normalized singular values,outperforms other conventional sparse measures such as the?_1 norm for network *** other is a spatial and temporal balance for tensorized neural *** accelerating the training of tensorized neural networks,it is crucial to leverage redundancy for both model compression and subspace training.

关键词： model compression subspace training effective rank low rank tensor optimization efficient deep learning

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

efficient Pose: efficient human pose estimation with neural architecture search

引用

Computational Visual Media 2021年第3期7卷 335-347页

作者： Wenqiang Zhang Jiemin Fang Xinggang Wang Wenyu Liu School of EIC Huazhong University of Science and TechnologyWuhan 430074China Institute of Artificial Intelligence Huazhong University of Science and TechnologyWuhan 430074China

Human pose estimation from image and video is a key task in many multimedia *** methods achieve great performance but rarely take efficiency into consideration,which makes it difficult to implement the networks on lightweight ***,real-time multimedia applications call for more efficient models for better ***,most deep neural networks for pose estimation directly reuse networks designed for image classification as the backbone,which are not optimized for the pose estimation *** this paper,we propose an efficient framework for human pose estimation with two parts,an efficient backbone and an efficient *** implementing a differentiable neural architecture search method,we customize the backbone network design for pose estimation,and reduce computational cost with negligible accuracy *** the efficient head,we slim the transposed convolutions and propose a spatial information correction module to promote the performance of the final *** experiments,we evaluate our networks on the MPII and COCO *** smallest model requires only0.65 GFLOPs with 88.1%PCKh@0.5 on MPII and our large model needs only 2 GFLOPs while its accuracy is competitive with the state-of-the-art large model,HRNet,which takes 9.5 GFLOPs.

关键词： pose estimation neural architecture search efficient deep learning

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

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

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

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：