Nelder-Mead algorithm
in [Matlab]
|
From: Jan on 7 Jun 2010 05:36 Thanks for your reply, it is now clear to me how the algorithm works. But, when the algorithm has found the maximum of a curve. How will it stop? Will the algorithm change the simplex a little until the maximum iterations are reached? Because when it has found the top, the optimalization paramters (Tol and TolX) will have no meaning. Or am I wrong? Thanks in advance Jan
From: John D'Errico on 7 Jun 2010 06:02 "Jan " <jan_neyens(a)skynet.be> wrote in message <huiei3$agl$1(a)fred.mathworks.com>... > Thanks for your reply, it is now clear to me how the algorithm works. > > But, when the algorithm has found the maximum of a curve. How will it stop? > Will the algorithm change the simplex a little until the maximum iterations are reached? > Because when it has found the top, the optimalization paramters (Tol and TolX) will have no meaning. Or am I wrong? > READ THE EXPLANATION OF NELDER-MEAD THAT I PROVIDED. It does explain this! As I said, the algorithm assumes that what goes down will continue to go down. Actually, I recall that Nelder-Mead moves away from high values in its attempt to minimize a function. However, if a new test point increases, it looks inside the simplex, changing the shape or size of the simplex as is appropriate. John
From: Jan on 7 Jun 2010 16:29 Sorry for my questions, but it is not really clear to me.. For example, I have a Gaussian curve. And a give an initial guess where this peak is. How does the algorithm search the peaks, searches the local maxima? Is the simplex used to search for the highest point in my signal? Thanks in advance Jan
From: Jan on 7 Jun 2010 16:43 They always say that the algorithm will go to a minimum. But, did they mean that it will go to a minium of (original data - estimated data)? Because I think I confusing the minimization of the function and searching the local maxima (peaks) Thanks in advance Jan
From: Alan Weiss on 8 Jun 2010 09:45
On 6/7/2010 4:43 PM, Jan wrote: > They always say that the algorithm will go to a minimum. But, did they > mean that it will go to a minium of (original data - estimated data)? > > Because I think I confusing the minimization of the function and > searching the local maxima (peaks) > > Thanks in advance > Jan There is a reasonably complete description of the algorithm in the documentation: http://www.mathworks.com/access/helpdesk/help/techdoc/math/bsgpq6p-11.html Alan Weiss MATLAB mathematical toolbox documentation |