Lsqnonlin with nonlinear constraint!

Question

GMark 2015-9-6

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/240350-lsqnonlin-with-nonlinear-constraint

评论： GMark 2015-9-8

Dear Community, I need to use a nonlinear constraint with the "lsqnonlin" optimization algorithm. To apply the constraint I am using the following procedure. FF = myfunction(pars). In myfunction I use the IF statement as follows: IF "nonlin constraints are respected" then compute FF ELSE FF=10e100 END. In other words, I make the objective function very large if the constraint is not respected, making the algorithm search for other parameters. Since I do not know how the algorithm works, is this a feasible solution? Thanks

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

Alan Weiss 2015-9-8

1
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/240350-lsqnonlin-with-nonlinear-constraint#answer_191665

It seems to me that you have a nonlinear constraint with a sum of squares problem. If you examine the Optimization Decision Table, you see that you should use fmincon, not lsqnonlin. Rewrite the objective function for fmincon so that it is a sum of squares, F'*F if F is a column vector of the components of the lsqnonlin objective function.

And no, it is not a good idea to have a non-smooth objective function that takes an arbitrary large value for infeasible points.

Alan Weiss

MATLAB mathematical toolbox documentation

1 个评论
显示 -1更早的评论隐藏 -1更早的评论

GMark 2015-9-8

Hi Alan, thanks a lot for responding. This is exactly the answer I wanted. I am already using fmincon and the on-website function fminsearchcon to double check my optimization. I wanted to use lsqnonlin as an extra tool, but I kind of realized it was not a good idea to break the function for non-feasible parameters. Thanks a lot

请先登录，再进行评论。