What kind of reason will cause failed after a long term of solved


(bloo) #1

It looks like this.


I have a function solved by CVX,I subtracted something from that and i think it will be also concave.
but i got a lot of situation like this,
however i can get correct result sometimes,There are about three times in 10 times,so i think my function is really concave.
so,can i get some information about it is “Can not find a feasible point“ or “Can not be solved in a limited iteration ”


(Mark L. Stone) #2

Read this material on CVX’s successive approximation method. http://cvxr.com/cvx/doc/advanced.html#the-successive-approximation-method It doesn’t always succeed. The fact that you are getting this output means CVX accepted your problem as complying with its rules.

You can try a different solver. Alternatively, you can use CVX 3.0 beta with a solver with which CVX will use native exponential cone capability instead of the successive approximation method. Specifically, ECOS, if you have no semidefinite constraints, or SCS (whether or not you have semidefinite constraints). SCS might be slow though, due to being a first order solver.

It;s also possible that your model formulation could be improved, which might help the solution process.


The procedure of CVX status show solved, while the final results status failed and Optimal value NAN?