Error using sqpLineSearch (line 22) Objective function is undefined at initial point. Fmincon cannot continue.

5 ビュー (過去 30 日間)
Hi,
I am trying to fit a new model in my data using fmincon:
[x,fval]=fmincon(@residualF,[0.01 0.01 0.01],[],[],[],[],[0 0 0],[10 10 10],[],options,Final_T(1:25),Int_A(1:25),5,Int_G(1,1:25))
Unfortunately Matlab gives the following error:
Error using sqpLineSearch (line 22)
Objective function is undefined at initial point. Fmincon cannot continue.
I used the same parameters before for other models. This new model is in frequency domain though so I use 'ifft' before I perform the optimization. I also checked that for my previous models in the frequency domain and it's working. Anyone would have any ideas why I have this error in my new model?
Thanks in advance!
G

採用された回答

Matt J
Matt J 2013 年 6 月 17 日
編集済み: Matt J 2013 年 6 月 17 日
If you execute
residualF([0.01 0.01 0.01])
you will see that it does not return a finite real value. Possibly, you haven't called ifft with the 'symmetric' option and it is returning a complex-valued result.
  2 件のコメント
George Papas
George Papas 2013 年 6 月 17 日
Thank you Matt. So, you would suggest to use ifftshift if I understood correctly? I know that this can slow down the optimization process by a lot though. I am open to ideas..
Matt J
Matt J 2013 年 6 月 18 日
No, I'm not sure where ifftshift would be relevant here. I was talking about the 'symmetric' option of ifft()

サインインしてコメントする。

その他の回答 (1 件)

George Papas
George Papas 2013 年 6 月 19 日
I used 'symmetric', still not working. I am afraid there is something wrong with the multiplication between the two functtions in the frequency domain (i.e. with my matrices). As you said Matt, I have no finite real values.
G

カテゴリ

Help Center および File ExchangeSurrogate Optimization についてさらに検索

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by