温州大学学报(自然科学版)
溫州大學學報(自然科學版)
온주대학학보(자연과학판)
JOURNAL OF WENZHOU UNIVERSITY(NATURAL SCIENCES)
2014年
3期
55-62
,共8页
边缘Fisher分析%类内紧凑图%类间分离图%主动学习%图像分类
邊緣Fisher分析%類內緊湊圖%類間分離圖%主動學習%圖像分類
변연Fisher분석%류내긴주도%류간분리도%주동학습%도상분류
Margin Fisher Analysis%Within-class Compact Graph%Inter-class Separate Graph%Active Learning%Image Classification
将边缘Fisher分析引入到MAED算法中,通过构建类内紧凑图和类间分离图,来描述样本点间的几何特征,形成一种新的主动学习方法。该算法利用两个图同时对流形数据局部结构和类鉴别信息进行建模,从而更好地保持了数据的内在几何特征。基于图像数据集的实验结果,证实了该方法的有效性。
將邊緣Fisher分析引入到MAED算法中,通過構建類內緊湊圖和類間分離圖,來描述樣本點間的幾何特徵,形成一種新的主動學習方法。該算法利用兩箇圖同時對流形數據跼部結構和類鑒彆信息進行建模,從而更好地保持瞭數據的內在幾何特徵。基于圖像數據集的實驗結果,證實瞭該方法的有效性。
장변연Fisher분석인입도MAED산법중,통과구건류내긴주도화류간분리도,래묘술양본점간적궤하특정,형성일충신적주동학습방법。해산법이용량개도동시대류형수거국부결구화류감별신식진행건모,종이경호지보지료수거적내재궤하특정。기우도상수거집적실험결과,증실료해방법적유효성。
In this paper, a new active learning method is put forward by leading margin Fisher Analysis into MAED algorithm, and describing the geometric structure of samples with constructing within-class compact and inter-class separate graphs. This algorithm uses two graphs modeling the local structure of the manifold data and discrimination information of the dataset simultaneously, which can keep the intrinsic geometric structure of the data better. This method is proved effectively based on the experimental results of the image data set.