为解决现有局部线性嵌入算法不适合处理非均匀分布数据和未利用距离远点信息的问题,提出相应的改进算法。首先引入测地线距离,以便能利用远点信息;然后使用调和平均规范化构造调和平均测地线核矩阵,使算法能更好地处理分布不均匀数据并具有鲁棒性。在UCI数据集上的实验结果表明,改进后的算法能够取得比局部线性嵌入算法更好的降维效果。 In order to solve the problem that the existing local linear embedding algorithm is not suitable for dealing with the non-uniform distribution data and the distance information of the unused distance, a corresponding improved algorithm is proposed. Firstly, the geodesic distance is introduced to make use of the far-point information. Then, the harmonic average geodesic kernel matrix is ​​constructed by harmonic normalization, which makes the algorithm better handle the unevenly distributed data and is robust. The experimental results on the UCI dataset show that the improved algorithm can achieve better dimensionality reduction than the local linear embedding algorithm.