In recent years, an LRLS algorithm which is highly accurate and stable in a more noisy environment than in LS one has been proposed. However, the computational complexity in the LRLS algorithm is on the order of n{sup}3. In this report, a fast LRLS algorithm whose computational complexity is on the order of n{sup}2 is considered. The fast LRLS algorithm consists of a method for the rank-one modification of the symmetric eigenproblem and an algorithm which updates an elgenvector on the order of n.
展开▼