0
$\begingroup$

I would like to extend my previous question What is difference between LMS and gradient-descent adaptation? with this other question.

I found out, that RLS and Kalman filter learning seems to be somehow similar. My question is: Can be those algorithms called gradient descent methods? If not, how is this kind of algorithms called?

Improve this question
$\endgroup$
3
  • 1
    $\begingroup$ Recursive least squares (RLS) filters don't use gradient descent. As their name suggests, they use a least-squares fit to determine the optimum coefficients at each time step. Via clever formulation of the filter structure, one can use the calculations done from time step $n$ to recursively calculate the updated coefficients for time step $n+1$ without having to do the full least-squares fit again. $\endgroup$
    – Jason R
    May 13, 2016 at 12:23
  • $\begingroup$ @JasonR Good to know. Does this approach a common name? Can you create an answer from your comment please? $\endgroup$
    – matousc
    May 23, 2016 at 9:27
  • $\begingroup$ Done. I'm not sure what you mean about the name of the approach; it's recursive least squares. $\endgroup$
    – Jason R
    May 23, 2016 at 12:50

1 Answer 1

Reset to default
2
$\begingroup$

Recursive least squares (RLS) filters don't use gradient descent. As their name suggests, they use a least-squares fit to determine the optimum coefficients at each time step. Via clever formulation of the filter structure, one can use the calculations done from time step $n$ to recursively calculate the updated coefficients for time step $n+1$ without having to do the full least-squares fit again.

Improve this answer
$\endgroup$
3
  • $\begingroup$ well, the least square solution actually nullify the gradient at each step. $\endgroup$
    – LJSilver
    May 23, 2016 at 13:53
  • $\begingroup$ //Via clever formulation of the filter structure,// You mean the lattice structure? $\endgroup$
    – Naveen
    May 23, 2016 at 15:08
  • $\begingroup$ Actually, this is far from optimal. And the $\lambda$ dependence turns it a parametric algorithm, such any standard gradient procedure. $\endgroup$
    – Brethlosze
    Jan 16, 2017 at 5:31

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.