南通大学学报(自然科学版)
南通大學學報(自然科學版)
남통대학학보(자연과학판)
JOURNAL OF NANTONG UNIVERSITY (NATURAL SCIENCE)
2013年
2期
57-61
,共5页
王雅玡%贾志洋%高炜
王雅玡%賈誌洋%高煒
왕아아%가지양%고위
半监督超图%正则拉普拉斯%分类算法%广义界
半鑑督超圖%正則拉普拉斯%分類算法%廣義界
반감독초도%정칙랍보랍사%분류산법%엄의계
semi-supervised hypergraph%normalization Laplacian%classification algorithm%generalization bound
研究超图上半监督正则拉普拉斯降维框架下分类算法的统计特征,通过特征空间上的映射算子说明降维后得到的向量较好地逼近原目标向量。利用矩阵迹的特性给出降维条件下分类算法广义界估计值。
研究超圖上半鑑督正則拉普拉斯降維框架下分類算法的統計特徵,通過特徵空間上的映射算子說明降維後得到的嚮量較好地逼近原目標嚮量。利用矩陣跡的特性給齣降維條件下分類算法廣義界估計值。
연구초도상반감독정칙랍보랍사강유광가하분류산법적통계특정,통과특정공간상적영사산자설명강유후득도적향량교호지핍근원목표향량。이용구진적적특성급출강유조건하분류산법엄의계고계치。
In this paper, the statistical characteristics of hypergraph classification algorithm under semi-supervised nor-malization laplacian dimension reduction were studied. By defining projection operator on the feature space, it is veri-fied that vector after dimension reduction well approximate to the original target vector. The generalization bound for classification algorithm under dimension reduction is estimated by using the trick of matrix trace.