what is better to use: fmincon OR lsqnonlin
in [Matlab]
|
Prev: Object Oriented Changes in 2008a
Next: need explanation: something that works with \ operator doesnt work with lsqnonlin function
From: kem on 5 May 2008 14:21 Hi, My objective function is to minimize the L2 norm of some vector + I have a linear constraint on the variables. So the output of the objfun: 1) in case of using lsqnonlin: F = [ f_1 ; f_2 ; .... ; f_n ; A* f - b] 2) and in case of fmincon the output is: sum ( f_i.^2 ) with the constraint A*f=b Which one of the two is better to use? what are the pros and cons of each of the methods? Thanks
From: Marcus M. Edvall on 5 May 2008 18:47
The best method for dense problems is NLSSOL in TOMLAB /NPSOL since you define the problem with residuals and linear constraints as one would normally want. This provides for better ways of scaling the problem and such. If the problem is sparse, then slsSolve in conjunction with TOMLAB / SNOPT is usually best since the residuals are transferred to the constraints and you can utilize the sparsity. Best wishes, Marcus Tomlab Optimization Inc. http://tomopt.com/tomlab/ On 5 mayo, 13:21, kem <keme...(a)gmail.com> wrote: > Hi, > > My objective function is to minimize the L2 norm of some vector + I > have a linear constraint on the variables. > > So the output of the objfun: > > 1) in case of using lsqnonlin: > F = [ f_1 ; f_2 ; .... ; f_n ; A* f - b] > > 2) and in case of fmincon the output is: > sum ( f_i.^2 ) with the constraint A*f=b > > Which one of the two is better to use? what are the pros and cons of > each of the methods? > > Thanks |