基于有监督流形学习的正交投影降维

α-based Supervised Orthogonal Projection Reduction by Affinity

将监督局部线性嵌入的思想引入传统的正交投影降维方法(OPRA)方法,提出一种新的基于有监督流形学习的正交投影降维方法(α-OPRA),使高维到低维的映射在保留某些流形结构的同时,进一步获得较好的正交投影效果。该方法通过加入额外的参数α来控制监督的程度,在纯粹的有监督的OPRA和无监督的OPRA之间取得了某些折中。实验结果证明,该方法能获得较好的降维结果。

This paper introduces the idea of SLLE into the traditional method of OPRA, which proposes a new approach of α-based Supervised Orthogonal Projection Reduction by Affinity（α-OPRA） for dimension reduction. Such method keeps the reservations of some flow-shaped structure during high-dimensional to low-dimensional mapping, gets better orthogonal projection. The method by adding additional parameters to control the degree of supervision, so in a purely supervised OPRA and unsupervised OPRA between there has been some compromise. Experimental results show that this method can get better reduction result.