Title :
New issues on the numerical stability of the fast recursive least squares algorithm
Author :
Gimenez, José Roberto B ; Romano, João Marcos T
Author_Institution :
Dept. of Commun., Univ. Estadual de Campinas, Sao Paulo, Brazil
Abstract :
The authors present a new model to describe the increase of the rounding noise in the fast recursive least squares algorithm when finite precision variables are used. The model is applied to the floating point representation approach and gives an approximate description of the noise accumulation process as it provides an accurate estimate of the divergence instant
Keywords :
filtering theory; floating point arithmetic; least squares approximations; noise; numerical stability; quantisation (signal); recursive estimation; LS algorithm; fast recursive least squares algorithm; finite precision variables; floating point representation; model; noise accumulation process; numerical stability; rounding noise; Adaptive filters; Electronic mail; Kalman filters; Least squares approximation; Least squares methods; Numerical stability; Phase estimation; Phase noise; Steady-state; Transversal filters;
Conference_Titel :
Circuits and Systems, 1995., Proceedings., Proceedings of the 38th Midwest Symposium on
Conference_Location :
Rio de Janeiro
Print_ISBN :
0-7803-2972-4
DOI :
10.1109/MWSCAS.1995.510259
