自适应邻域选择的I somap算法

Dimensionality reduction method based on adaptive neighborhood selection

针对流形学习算法Isomap对于稀疏数据局部邻域大小选择的敏感性,提出一种自适应邻域选择的降维方法 A-Isomap(Adaptive-Isomap).在数据稀疏的情况下,通过邻域选取算法自适应的动态选择每一个样本点的邻域大小,很大程度上避免了对短路点的选择;同时,使用聚类信息来汇聚相似的样本点,保证了降维后的数据具有很好的可分性.为了验证算法的有效性,将该算法应用于手工流形的降维,结果表明该算法能较好的展现降维效果.

The Isomap method based on manifold learning are sensitive to the selection of local neigh-bors.This paper proposed a dimensionality reduction based on adaptive neighborhood selection.In the case of sparse data sets,it select the neighborhood of each sample point by adaptive neighbor-hood selection algorithm,and avoid the the short-circuit point selection largely.While,it clustered the similar sample points by using clustering information,which guaranteed good separability for the reduction data.Located in hand manifolds for the high-dimensional data on the experiment to test the improved algorithm has a good effort of reducing dimension.