19-06-2012, 04:07 PM
RECURSIVE LEAST SQUARE ALGORITHM
RECURSIVE LEAST SQUARE ALGORITHM.ppt (Size: 933.5 KB / Downloads: 36)
RLSA
The calculation of Wm is time consuming & is not suitable for real time or online filtering. This method is suitable for real time filtering.
We use recursive methods to improve the estimate of Wm using new data.
With RLSA method the estimates of Wm can be updated for each set of new data acquired with out repeatedly solving the time consuming inversion matrix directly.
RLS algorithm can be obtained by increasing the power of the data to remove the effects of the old data.
It helps in tracking the slowly varying characteristics of a signal.
LIMITATIONS OF RLSA
RLS method is very efficient & involves exactly the same number of arithmetic operations between samples & have fixed dimensions.
problems in this method:
Blow up
Sensitivity to computer round off errors.
Second problem is due to its sensitivity to computer roundoff errors, which results in negative definite P matrix & this leads to instability.
This problem is more worse in multiparameter models, especially if the variables are linearly dependent & when the algorithm is implemented on a small system with finite wordlength.
The problem of numerical instability may be solved by suitably factorizing the matrix P such that the differencing terms are avoided.