OPT OpenIR  > 光学影像学习与分析中心
Cascaded Subpatch Networks for Effective CNNs
Jiang, Xiaoheng1; Pang, Yanwei1; Sun, Manli1; Li, Xuelong2; Pang, YW (reprint author), Tianjin Univ, Sch Elect & Informat Enginnering, Tianjin 300072, Peoples R China.
作者部门光学影像学习与分析中心
2018-07
发表期刊IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
ISSN2162-237X
卷号29期号:7页码:2684-2694
产权排序2
摘要

Conventional convolutional neural networks use either a linear or a nonlinear filter to extract features from an image patch (region) of spatial size H x W (typically, H is small and is equal to W, e.g., H is 5 or 7). Generally, the size of the filter is equal to the size H x W of the input patch. We argue that the representational ability of equal-size strategy is not strong enough. To overcome the drawback, we propose to use subpatch filter whose spatial size h x w is smaller than H x W. The proposed subpatch filter consists of two subsequent filters. The first one is a linear filter of spatial size h x w and is aimed at extracting features from spatial domain. The second one is of spatial size 1 x 1 and is used for strengthening the connection between different input feature channels and for reducing the number of parameters. The subpatch filter convolves with the input patch and the resulting network is called a subpatch network. Taking the output of one subpatch network as input, we further repeat constructing subpatch networks until the output contains only one neuron in spatial domain. These subpatch networks form a new network called the cascaded subpatch network (CSNet). The feature layer generated by CSNet is called the csconv layer. For the whole input image, we construct a deep neural network by stacking a sequence of csconv layers. Experimental results on five benchmark data sets demonstrate the effectiveness and compactness of the proposed CSNet. For example, our CSNet reaches a test error of 5.68% on the CIFAR10 data set without model averaging. To the best of our knowledge, this is the best result ever obtained on the CIFAR10 data set.

关键词Cascaded Subpatch Networks (Csnet) Convolutional Neural Network (Cnn) Feature Extraction Subpatch Filter
学科领域Computer Science, Artificial Intelligence
DOI10.1109/TNNLS.2017.2689098
收录类别SCI
语种英语
WOS研究方向Computer Science ; Engineering
WOS记录号WOS:000436420400002
引用统计
被引频次:2[WOS]   [WOS记录]     [WOS相关记录]
文献类型期刊论文
条目标识符http://ir.opt.ac.cn/handle/181661/30415
专题光学影像学习与分析中心
通讯作者Pang, YW (reprint author), Tianjin Univ, Sch Elect & Informat Enginnering, Tianjin 300072, Peoples R China.
作者单位1.Tianjin Univ, Sch Elect & Informat Enginnering, Tianjin 300072, Peoples R China
2.Chinese Acad Sci, Xian Inst Opt & Precis Mech, State Key Lab Transient Opt & Photon, Ctr OPT IMagery Anal & Learning, Xian 710119, Shaanxi, Peoples R China
推荐引用方式
GB/T 7714
Jiang, Xiaoheng,Pang, Yanwei,Sun, Manli,et al. Cascaded Subpatch Networks for Effective CNNs[J]. IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS,2018,29(7):2684-2694.
APA Jiang, Xiaoheng,Pang, Yanwei,Sun, Manli,Li, Xuelong,&Pang, YW .(2018).Cascaded Subpatch Networks for Effective CNNs.IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS,29(7),2684-2694.
MLA Jiang, Xiaoheng,et al."Cascaded Subpatch Networks for Effective CNNs".IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS 29.7(2018):2684-2694.
条目包含的文件
文件名称/大小 文献类型 版本类型 开放类型 使用许可
Cascaded Subpatch Ne(2610KB)期刊论文作者接受稿开放获取CC BY-NC-SA浏览 请求全文
个性服务
推荐该条目
保存到收藏夹
查看访问统计
导出为Endnote文件
谷歌学术
谷歌学术中相似的文章
[Jiang, Xiaoheng]的文章
[Pang, Yanwei]的文章
[Sun, Manli]的文章
百度学术
百度学术中相似的文章
[Jiang, Xiaoheng]的文章
[Pang, Yanwei]的文章
[Sun, Manli]的文章
必应学术
必应学术中相似的文章
[Jiang, Xiaoheng]的文章
[Pang, Yanwei]的文章
[Sun, Manli]的文章
相关权益政策
暂无数据
收藏/分享
文件名: Cascaded Subpatch Networks for Effective CNNs.pdf
格式: Adobe PDF
所有评论 (0)
暂无评论
 

除非特别说明,本系统中所有内容都受版权保护,并保留所有权利。