基于储备池计算网络的小样本图像分类方法  

Reservoir computing based network for few-shot image classification

作　　者：王彬[1] 兰海[2] 俞辉 郭杰龙 魏宪 WANG Bin;LAN Hai;YU Hui;GUO Jie-long;WEI Xian(School of Advanced Manufacturing,Fuzhou University,Quanzhou 362200,China;Fujian Institute of Research on the Structure of Matter,Chinese Academy of Sciences,Fuzhou 350002,China;Fujian Science&Technology Innovation Laboratory for Optoelectronic Information of China(Mindu Innovation Laboratory),Fuzhou 350108,China)

机构地区：[1]福州大学先进制造学院,福建泉州362200 [2]中国科学院福建物质结构研究所,福建福州350002 [3]中国福建光电信息科学与技术创新实验室(闽都创新实验室),福建福州350108

出　　处：《液晶与显示》2023年第10期1399-1408,共10页Chinese Journal of Liquid Crystals and Displays

基　　金：中国福建光电信息科学技术创新实验室(闽都创新实验室)(No.2021ZZ120);福建省科技项目计划(No.2021T3003);泉州市科技项目(No.2021C065L)。

摘　　要：针对目前小样本学习方法易过拟合、跨域泛化能力不足等问题,受启发于储备池计算不依赖于训练而缓解过拟合的特性,提出了一种基于储备池计算的小样本学习方法(Reservoir Computing based Network for Few-shot Image Classification,RCFIC)。整个方法由特征提取模块、特征增强模块和分类器模块构成。特征增强模块由储备池模块和基于储备池的注意力机制构成,分别对特征提取网络的特征进行通道级增强和像素级增强,同时联合余弦分类器促使网络学习具有高类间方差、低类内方差特性的特征分布。实验结果表明,本文算法在Cifar-FS、FC100、Mini-ImageNet等数据集上的分类精度至少比现有方法高1.07%,在从Mini-ImageNet到CUB-200的跨域场景设置下的分类精度优于次优方法1.77%。同时,消融实验验证了RCFIC的有效性。所提方法泛化性强,能够有效缓解小样本图像分类中的过拟合问题并在一定程度上解决跨域问题。Aiming at the problems that current few-shot learning algorithms are prone to overfitting and insufficient generalization ability for cross-domain cases,and inspired by the property that reservoir computing(RC)does not depend on training to alleviate overfitting,a few-shot image classification method based on reservoir computing(RCFIC)is proposed.The whole method consists of a feature extraction module,a feature enhancement module and a classifier module.The feature enhancement module consists of a RC module and an attention mechanism based on the RC,which performs channel-level enhancement and pixel-level enhancement of the features of the feature extraction module,respectively.Meanwhile,the joint cosine classifier drives the network to learn feature distributions with high inter-class variance and low intra-class variance properties.Experimental results indicate that the algorithm achieves at least 1.07%higher classification accuracy than the existing methods in Cifar-FS,FC100 and Mini-ImageNet datasets,and outperforms the second-best method in cross-domain scenes from Mini-ImageNet to CUB-200 by at least 1.77%.Meanwhile,the ablation experiments verify the effectiveness of RCFIC.The proposed method has great generalization ability and can effectively alleviate the overfitting problem in few-shot image classification and solve the cross-domain problem to a certain extent.

关 键 词：小样本学习 储备池计算 注意力机制 特征增强 图像分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]