论文部分内容阅读
提出一种基于局部线性判别器融合的方法,在非线性流形上展开判别分析.首先根据 Gabriel 图对整体流形作局部区域划分,并构造局部线性判别器.然后通过局部判别器融合获取整体非线性判别器:基于边界准则函数,以迭代优化的方式为每个局部判别器分配最佳的权重系数.基于边界准则函数的融合算法,克服小样本问题,消除整体判别器的性能对样本分布的依赖性.在人工合成数据集以及人脸图像库上的实验证明本文算法的有效性.
This paper presents a method based on local linear discriminator fusion, which develops discriminant analysis on nonlinear manifolds. Firstly, the local manifolds are partitioned according to Gabriel graphs and local linear discriminators are constructed, then the whole discriminator is obtained through local discriminator fusion Nonlinear discriminator: Based on the boundary criterion function, each local discriminator is assigned an optimal weight coefficient in an iterative optimization way.A fusion algorithm based on the boundary criteria function is used to overcome the small sample problem and eliminate the performance of the overall discriminator against the sample distribution The experiments on synthetic datasets and face images demonstrate the effectiveness of our algorithm.