解张量最佳秩一逼近问题一些方法.pdf

21 ALS Jacobi Gauss-Newton GRQ-ALS GRQ-Jacobi GRQ-Jacobi Lagrange AGS GRQ-Newton NGRQ- – v – Newton – vi – Abstract Abstract In the twenty-first century, many problems about tensors rise from the area of Psychology, Economy, Chemistry, Sensor Array Processing and so on. As a result, tensor computation is paid on more and more attentions. Tensor decomposition and approximation are widely applied in High Order Statistics and Signal Processing. First, we give a brief review of the usual methods for rank-one ap- proximation to higher-order tensors. Second, the ALS and Jacobi-Gauss-Newton methods for rank-one approximation are usually used. However, their convergence rate is not ideal. By analysis and comparison, we obtain GRQ-ALS method and GRQ-Jacobi method which have smaller approximation error and fast convergence rate. We also give the global convergence proof for GRQ- Jacobi method. Third, we