fminsearchbnd optimization function in matlab

Question

mohammed hussein 2015-12-2

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/258471-fminsearchbnd-optimization-function-in-matlab

回答： MASTER WHOS 2019-2-8

i used fminsearchbnd for boundary problem but the matlab it tell me that Undefined function 'fminsearchbnd' can any one help me with is problem i used this example

fun = @(x)(x(1)-5)^2+(x(2)+6)^2+x(3); % x0 = [0 0 0]; Lb=[0 -1 2]; Ub=[1 2 3]; [x,fval,exitflag,output] = fminsearchbnd(fun,Lb,Ub); disp(fval)

thank you

0 个评论
显示 -2更早的评论隐藏 -2更早的评论

请先登录，再进行评论。

请先登录，再回答此问题。

Answer 1

Walter Roberson 2015-12-2

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/258471-fminsearchbnd-optimization-function-in-matlab#answer_201871

fminsearchbnd is not part of MATLAB. You can get it one of the versions from the File Exchange

3 个评论
显示 1更早的评论隐藏 1更早的评论

Walter Roberson 2015-12-2

Do not put it in the MATLAB toolbox. Download the .zip file, and unzip it into a directory that has nothing else in it (and which is not under the MATLAB installation.) Then use pathtool to add the directory to the MATLAB path. After that MATLAB will be able to find the file.

mohammed hussein 2015-12-2

Thank you very much

请先登录，再进行评论。

Answer 2

MASTER WHOS 2019-2-7

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/258471-fminsearchbnd-optimization-function-in-matlab#answer_359916

I DID THIS:

Do not put it in the MATLAB toolbox. Download the .zip file, and unzip it into a directory that has nothing else in it (and which is not under the MATLAB installation.) Then use pathtool to add the directory to the MATLAB path. After that MATLAB will be able to find the file.

but it gives error: Error in fminsearchbnd (line 264)

[xu,fval,exitflag,output] = fminsearch(intrafun,x0u,options,varargin);

what might be the problem?

11 个评论
显示 9更早的评论隐藏 9更早的评论

John D'Errico 2019-2-8

Ok. Your very first problem is that you cannot realistically use fminserchbnd with any serious hope of success.

You have 56 unknowns to estimate. Although one of them seems to have a lower bound of 1 which is equal to its upper bound of 1. The latter is not a problem though. Fminserchbnd can handle that case. In fact, I recall that I specifally catered to that case, looking for variables that had lower and upper bounds that were identical. So really, you have 55 unknowns to estimate.

Regardelss, that is by far too many unknowns for fminsearchbnd to handle. Perhaps 10-12 unknowns is a far more reasonable limit to gain convergence from a tool like fminsearchbnd. Once you start to try to solve problems with many tens of parameters like this, tools like fminsearch and fminsearchbnd tend to get bogged down, becoming far too inefficient. fminsearchbnd would probably require literally billions of function evaluations to achieve even approximate convergence.

I would strongly recommend instead using fmincon from the optimization toolbox, if that is an option.

(And please stop adding answers just to add comments.)

Torsten 2019-2-8

I still don't see neither

@SQP_complex2

nor

@SQP_complex2_case2

Walter Roberson 2019-2-8

I have my own private implementation of tricking fminsearch into being bounded (stops when it hits the boundary rather than reflecting as might be nice.)

I know from experience that 55 variables is completely intractable to use seriously with fminsearch . fminsearch gets caught in local minima , not as easily as fmincon gets caught but it happens all the time. It is not a global minimizer. To use it for more than light use you need to start it at multiple locations . In particular you need starting points near every valid combination of upper and lower bounds . With 55 variables that would be 2^55 configuration to examine. If you could process 2^30 per second (gigaflop range ) then that would take you 2^25 seconds . That would be a year . You probably cannot process more than 1/100 of that speed so probably 100 or so years . Just for figuring out crudely which end of each bound you probably near.

请先登录，再进行评论。