Study on noise reduction in singular value decomposition based on structural risk minimization
Yungong Li · 2005
The order of the effective rank of a matrix is difficult to determine for noise reduction in singular value decomposition. A method is presented for solving this problem based on structural risk minimization. Based on statistical learning theory, the determination of the order of an effective rank is considered a learning process. The order of the effective rank can be achieved by using the structural risk minimization(SRM) instead of the experience risk minimization(ERM). Simulation results show that this method has good noise reduction accuracy and deduces the complexity of the algorithm.