fminicon optimization function and NaN problem

Question

sal 2017-8-28

0
链接

此问题的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/354319-fminicon-optimization-function-and-nan-problem

评论： sal 2017-8-28

fminicon optimization and NaN problem ,

I have a function that i am trying to minimize it using the fminicon optimization function and the results return are always NaN, i have debugged the program and checked the reason of the NaN it was because i have some results in the function equal to zero then in the function this results should take log and it will give -inf the when multiplied by zero it will give NaN,

my question is that can i do anything to make the fminicon function overcome the problem of the NaN?

i would really appreciate your help

thanks in advance

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 2017-8-28

0
链接

此回答的直接链接

https://ww2.mathworks.cn/matlabcentral/answers/354319-fminicon-optimization-function-and-nan-problem#answer_279463

If you give a starting point that is feasible (no NaN values), and use the interior-point algorithm, then fmincon should be able to handle the issue.

You could also write your objective function to explicitly look for the issue and return proper values.

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

sal 2017-8-28

Thank you so much for your reply, Regarding my initial it is feasible because it comes from a searching algorithm.

and for the rewriting the objective function i don't have an idea about how to write the objective function in a way that when it find (log(0)) turn it into a very small value not -inf to avoid the problem in the next step of multiplying the -inf by 0 which produces the NaN problem,

I wonder if there is anyway to put such condition in the objective function code since i have tried multiple ways but still none o them worked.

i would really appreciate if you suggest some way to deal with this issue.

请先登录，再进行评论。