论文部分内容阅读
局部线性判别嵌入(locally linear discriminant embedding,LLDE)将局部线性嵌入(locally linear embedding,LLE)和最大间隔(maximum margin criterion,MMC)进行融合,有效地提高了LLE算法的识别力。但其保留的是数据的全局判别信息,且依赖数据的分布。针对LLDE的不足,本研究将LLE和加权非参数最大间隔(weighted non-parametric maximum margin criterion,WNMMC)进行融合,提出了一种新的有监督的降维方法——非参数判别性局部线性嵌入(nonparametric locally linear discriminant embedding,NLLDE)。NLLDE保留了数据更为有效的局部判别信息,因此更具判别力。NLLDE采用了非参数数据表示,使得模型及求解不依赖于数据的分布,克服了LLDE针对高斯分布数据有效的局限,其应用范围更为广泛。Yale和PIE人脸数据库上的实验结果证实了NLLDE的高效性。
Locally linear discriminant embedding (LLDE) merges local linear embedding (LLE) and maximum margin criterion (MMC), effectively improving the recognition ability of LLE algorithm. But it retains the data of the global discriminant information, and depends on the distribution of data. In order to overcome the deficiency of LLDE, this paper merges LLE with weighted non-parametric maximum margin criterion (WNMMC), and proposes a new supervised dimensionality reduction method - nonparametric discriminative local linear embedding (nonparametric locally linear discriminant embedding, NLLDE). NLLDE retains the data more effective local discriminant information, so more discriminatory. NLLDE adopts nonparametric data representation, which makes the model and solution independent of data distribution and overcomes the limitation of LLDE for Gaussian distribution data. Its application is more extensive. The experimental results on the Yale and PIE face databases confirm the efficiency of NLLDE.