convexity checked failure using ipopt

[yongjie wang]

Hello,Johan

When i run a program,it shows    EXIT: Converged to a point of local infeasibility. Problem may be infeasible.it may be a bad model,in my model, it contains matlab command   conj and real-imag of complex variables such as [P+Q*i==U.*(conj(jiedianY*U))];  i is sqrt(-1), and jiedianY is a matrix consists of complex numbers.However U is a vector variables and the variables are complex varaibles =Ue+Uf*i,and Ue and Uf are real variables.can ipopt check it,or it violates its checking rules?

thanks,i do not know the reason,and debug maybe do not work.

[Johan Löfberg]

I am not sure what you are asking. ipopt never sees the conj/real/imag commands, those are applied at the model level by YALMIP to manipulate the real variables which define the complex number.

An obvious problem with your model is that it contains a quadratic equality, i.e., it is nonconvex. Although ipopt accepts nonconvex problems, you are not guaranteed that it will work well.

[yongjie wang]

Thanks for your answer,then i try to change the express way,and reform it to  [P==Ue.*(chaifenshi*Ue-chaifenxu*Uf)+Uf.*(chaifenshi*Uf+chaifenxu*Ue)],and the variables are all real variables and chaifenshi----chaifenxu are real numbers matrix,and others are real vector variables,it turns out 

Maximum iterations or time limit exceeded (IPOPT)

I know it exceed the max limitation,but how can i deal with it or set which parameters to let it run?

Thanks

[Johan Löfberg]

Problem is still nonconvex.

The available options are created, as always, using sdpsettings
ops=sdpsettings;ops.ipopt

[yongjie wang]

But i do not know which one should i set,i open the ops.ipopt,but i can not find the upper  limitations in it .

[Johan Löfberg]

Hmm, that option does not appear to be available through the options structure I create through OPTI Toolbox function ipoptset. I will have to ask Jonathan about it.

However, you should not try to fix the symptom, but the underlying problem. The fact that you hit the iteration limit indicates that you have a bad model. What is your model?

[yongjie wang]

As the constraints,it is a very copmlex model,and contains quadratic nonconvex equations.Thanks

[Johan Löfberg]

Can you send it to me for analysis

[Johan Löfberg]

You'' have to add it manually

ops.ipopt.max_iter = 12345;

But as I said, your problem is probably not the iteration limitation but properties of your model