Adaptive RRQR-based factorization: Improving the algorithm tracking capabilities
Monique P. Fargues, M.P. Ferreira · 2003
An improvement to a rank-revealing QR (RRQR) updating procedure proposed earlier is presented. The method takes advantage of the simplicity of the QR updating scheme, the rank-revealing property proposed by T.F. Chan (1987), and the Hermitian property of correlation matrices. Simulations show that the 'refined' RRQR-based updating scheme approximates signal (and noise) subspaces closely and is robust to noise power errors.>