基于局部线形嵌入算法的雷达目标识别

维数约减问题出现在信息处理的许多方面,非线性方法主要有局部线性嵌入LLE (Locally Linear Embedding) 、拉普拉斯特征映射(Laplacian Eigenmap) 、基于Hessian 矩阵的LLE 等,它们通过在高维空间中设计数据集所在流形的拓扑、几何等特性,很好地弥补了线性降维不能发现数据集非线性结构的不足.其中局部线性嵌入这种非监督学习算法应用广泛,在此基础上将其用于作为雷达目标识别的五种飞机数据,取得了很好的效果.

Radar Target Recognition Based on Locally Linear Embedding

The problem of dimension reduction arises in many fields of information processing, including machine leaning, data compression, pattern recognition. In this paper, the author describes locally linear embedding, an unsupervised leanting algorithm and the details parameters of the algorithm, and discusses several extensions. The result of the algorithm is applied to the data of five typical planes for radar target recognition, which has good impact.