CAJ | 학술논문

作为一种降维的方法，稀疏编码的有效性很高。编码字典具有超完备性，在编码中会丢失不同编码字典特征的衔接信息与相似度，因而使得稀疏编码自身的识别率有所下降。针对编码字典不同特征的距离与相似度，本文构建了新型的超图稀疏编码框架，在计算稀疏编码中，可以在此结构基础上运用相似度权重和超图理论，有效的改善了稀疏编码方法，使稀疏编码特有的鲁棒性得到增强。接着分别在Caltech和Scene数据库做实验，对该稀疏编码的有效性进行了验证。
작위일충강유적방법，희소편마적유효성흔고。편마자전구유초완비성，재편마중회주실불동편마자전특정적함접신식여상사도，인이사득희소편마자신적식별솔유소하강。침대편마자전불동특정적거리여상사도，본문구건료신형적초도희소편마광가，재계산희소편마중，가이재차결구기출상운용상사도권중화초도이론，유효적개선료희소편마방법，사희소편마특유적로봉성득도증강。접착분별재Caltech화Scene수거고주실험，대해희소편마적유효성진행료험증。
As a method of dimension reduction, effective sparse coding is very high. The encoding dictionary has super completeness, link information and similarity can be lost in different coding dictionary features in the code, which makes the identification of sparse coding itself has fallen. According to the distance and similarity coding dic-tionary with different characteristics, we constructs a hypergraph sparse coding framework model, in the calculation of sparse coding, can be based on the use of structural similarity weight and hypergraph theory, effectively improve the sparse coding method, the sparse coding features some robustness is enhanced. Then do the experiments on Caltech and Scene database, the validity of the sparse coding is verified.