典型文献
Sharing Weights in Shallow Layers via Rotation Group Equivariant Convolutions
文献摘要：
The convolution operation possesses the characteristic of translation group equivariance.To achieve more group equivari-ances,rotation group equivariant convolutions(RGEC)are proposed to acquire both translation and rotation group equivariances.However,previous work paid more attention to the number of parameters and usually ignored other resource costs.In this paper,we construct our networks without introducing extra resource costs.Specifically,a convolution kernel is rotated to different orientations for feature extractions of multiple channels.Meanwhile,much fewer kernels than previous works are used to ensure that the output channel does not increase.To further enhance the orthogonality of kernels in different orientations,we construct the non-maximum-suppression loss on the rotation dimension to suppress the other directions except the most activated one.Considering that the low-level-features be-nefit more from the rotational symmetry,we only share weights in the shallow layers(SWSL)via RGEC.Extensive experiments on mul-tiple datasets(i.e.,ImageNet,CIFAR,and MNIST)demonstrate that SWSL can effectively benefit from the higher-degree weight shar-ing and improve the performances of various networks,including plain and ResNet architectures.Meanwhile,the convolutional kernels and parameters are much fewer(e.g.,75％,87.5％fewer)in the shallow layers,and no extra computation costs are introduced.
文献关键词：
中图分类号：
作者姓名：
Zhiqiang Chen;Ting-Bing Xu;Jinpeng Li;Huiguang He
作者机构：
Research Center for Brain-inspired Intelligence,Institute of Automation,Chinese Academy of Sciences,Beijing 100190,China;School of Instrumentation and Optoelectronic Engineering,Beihang University,Beijing 100191,China;Ningbo HwaMei Hospital,University of Chinese Academy of Sciences,Ningbo 315012,China;Center for Excellence in Brain Science and Intelligence Technology,Chinese Academy of Sciences,Beijing 100190,China
文献出处：
引用格式：
[1]Zhiqiang Chen;Ting-Bing Xu;Jinpeng Li;Huiguang He-.Sharing Weights in Shallow Layers via Rotation Group Equivariant Convolutions)[J].机器智能研究（英文）,2022(02):115-126
A类：
Equivariant,equivariance,equivari,RGEC,equivariances,extractions,nefit,SWSL,shar
B类：
Sharing,Weights,Shallow,Layers,via,Rotation,Group,Convolutions,operation,possesses,characteristic,translation,group,To,achieve,more,equivariant,convolutions,proposed,acquire,both,However,previous,paid,attention,number,parameters,usually,ignored,other,resource,costs,In,this,paper,construct,networks,without,introducing,Specifically,rotated,different,orientations,multiple,channels,Meanwhile,much,fewer,kernels,than,used,ensure,that,output,does,not,increase,further,enhance,orthogonality,maximum,suppression,loss,dimension,directions,except,most,activated,one,Considering,level,features,from,rotational,symmetry,only,share,weights,shallow,layers,Extensive,experiments,datasets,ImageNet,CIFAR,MNIST,demonstrate,can,effectively,benefit,higher,degree,improve,performances,various,including,plain,ResNet,architectures,convolutional,computation,introduced
AB值：
0.51435
相似文献
机标中图分类号，由域田数据科技根据网络公开资料自动分析生成，仅供学习研究参考。
