摘要
流形排序算法被广泛地应用到半监督学习领域中,然而其性能紧紧依赖于底层图结构。针对现有的流形排序算法效果欠佳的现状,提出了一种全新的图结构——自然邻居图,这种图能自适应流形结构,并且构造这种图不需要提前指定参数k。将这种图结构应用到基于流形排序的图像检索框架下,并证明了基于提出的自然邻居图的流形排序算法的效果优于基于KNN的流形排序算法。
The manifold-ranking based method is widely used in semi-supervised learning,and its performance is closely related to the structure of the constructed graph. To improve the performance of existing manifold-ranking based method,this paper presented a novel graph structure named natural neighbor graph and its corresponding constructing algorithm. However,there＇s no need to specify the free parameter k. It applied the new graph structure into the framework of manifold-ranking based image retrieval and showed that the manifold ranking algorithm based on graph structure performed better than KNN graph. Experiments demonstrate that the proposed algorithm outperforms the state-of-the-art algorithms.
出处
《计算机应用研究》
CSCD
北大核心
2016年第4期1265-1268,1276,共5页
Application Research of Computers
基金
国家自然科学基金资助项目(61272194)
重庆市自然科学基金资助项目(cstc2013jcyj A40049)
关键词
流形排序
自然邻居
图像检索
KNN
manifold-ranking
natural neighbor
image retrieval
KNN
