论文部分内容阅读
主要研究半监督局部线性嵌入算法(Semi-Supervised Locally Linear Embedding,简称SSLLE)对于噪声的敏感性,提出一种具有鲁棒性的半监督局部线性嵌入算法(Robust Semi-Supervised Locally Linear Embedding,简称RSSLLE).RSSLLE在对数据进行离群点检测的基础上,从两方面增加算法对离群点的鲁棒性.对于光滑点集,直接对其采用SSLLE算法进行降维,以避免离群点对光滑点的影响;对于离群点集,利用其局部投影坐标计算局部重构权,从而真正反映离群点的局部线性关系.再将光滑点集作为训练点集,结合SSLLE方法计算离群点集的低维坐标.模拟实验和实际例子表明RSSLLE对噪声有很好的鲁棒性.
This paper mainly studies the semi-supervised Locally Linear Embedding (SSLLE) sensitivity to noise and proposes a Robust Semi-Supervised Locally Linear Embedding (RSSLLE) ) RSSLLE increases the robustness of the algorithm to outliers based on outlier detection.For the smooth point sets, the SSLLE algorithm is directly used to reduce the dimension in order to avoid outlier pairs Smooth and smooth points.For the outlier sets, the local reconstruction weights are calculated by using the local projection coordinates, which can truly reflect the local linear relationship of the outliers.Secondly, we use the smooth point set as the training point set and the SSLLE method to calculate the outliers Set of low-dimensional coordinates.The simulation experiments and practical examples show that RSSLLE is robust to noise.