在许多应用中,都牵涉到对Kernel矩阵进行特征分解,同时矩阵分解的时间复杂度与矩阵的维数成正比。而在许多应用中,被分解的Kernel矩阵始终保持有低秩特性,因此基于Nystrom低秩逼近分解算法求取非线性子空间的算法通常被用来加速分解过程。提出一种增量Nystrom算法用于动态学习并将其应用动态跟踪问题。 In many applications, the decomposition of Kernel matrix is ​​involved, and the time complexity of matrix decomposition is proportional to the dimension of matrix. In many applications, the decomposition of the Kernel matrix always has a low rank, so the algorithm based on the Nystrom low rank approximation decomposition algorithm to obtain nonlinear subspaces is usually used to speed up the decomposition process. An incremental Nystrom algorithm is proposed for dynamic learning and its application to dynamic tracking.