基于自适应最近邻的局部线性嵌入算法

Locally Linear Embedding Algorithm Based on Adaptive Nearest Neighbor

局部线性嵌入算法是一个优异的非线性维数约减方法,但是算法本身是一个无监督学习算法,对于有监督问题的学习效果不是很好。这主要是因为算法使用了K-近邻方法来求解最近邻点。针对这个缺点,提出了一种改进的、基于自适应最近邻法的局部线性嵌入方法,数值实验证明算法对于有监督的学习问题,具有较好的适应性。

Locally linear embedding is an efficient nonlinear dimensional reduction algorithm. Because the algorithm is an unsupervised learning algorithm, the effect that deal with the supervised learning problem is no good. The reason is that the algorithm searches the nearest neighbor points with K-nearest neighbor. An adaptive nearest neighbor locally linear embedding algorithm is proposed to overcome this shortage. Experiment results show that the algorithm adapts well the supervised learning problems.