Undefined function or variable 'y_s' about sedumi HOT 22 CLOSED

So does SeDuMI really work on your machine, i.e. what happens if you run

>> sedumi(1,1,1)

from sedumi.

SeDuMi 1.3.4 by AdvOL, 2005-2008 and Jos F. Sturm, 1998-2003.
Alg = 2: xz-corrector, Adaptive Step-Differentiation, theta = 0.250, beta = 0.500
eqs m = 1, order n = 2, dim = 2, blocks = 1
nnz(A) = 1 + 0, nnz(ADA) = 1, nnz(L) = 1
it : by gap delta rate t/tP t/tD* feas cg cg prec
0 : 8.00E+00 0.000
1 : 7.87E-01 5.00E-01 0.000 0.0625 0.9900 0.9900 1.40 1 1 2.9E-01
2 : 1.00E+00 1.65E-04 0.000 0.0003 0.9999 0.9999 1.09 1 1
iter seconds digits cx by
2 0.2 Inf 1.0000000000e+00 1.0000000000e+00
|Ax-b| = 0.0e+00, [Ay-c]_+ = 0.0E+00, |x|= 1.0e+00, |y|= 1.0e+00

Detailed timing (sec)
Pre IPM Post
4.500E-02 1.050E-01 6.991E-03
Max-norms: ||b||=1, ||c|| = 1,
Cholesky |add|=0, |skip| = 0, ||L.L|| = 1.

ans =

 1

from sedumi.

This is the result, and I think it works well.

from sedumi.

Then you have to turn on debug in YALMIP in sdpsettings when making the call to see the crash in sedumi

from sedumi.

I broke into the sedumi. It seems that the function sedumi does not return the result 'y_s'. Then as putting the value of y_s to Primal, the error occurs, just like what the error message shows. But I didn't find what's wrong in the function sedumi. Maybe it is due to the size of the problem overweighs the ability of sedumi. Actuallay, under the same settings, the sedumi and yalmip work well now with a smaller prolem.

from sedumi.

The only reasons I ever see for crashes are

1. Wrong binary (which simple means it doesn't run)
2. Too trivial problem
3. Out of memory (is typically printed)

Trying another solver to see if there is something generically wrong with the model is always recommended

from sedumi.

In fact, I also tried gurobi and cplex with yalmip in matlab to solve the same problem. There is no error generated but a NaN result is ouput.

from sedumi.

NaN means the problem wasn't solved, so you probably had a diagnostic flag telling you some problem (unbounded or infeasible), so there is something weird/trivially wrong with your model that you have to sort out

from sedumi.

Exactly! The guroti said it can not solve a problem with semidefine constraint. But, I used sedumi solving a similiar problem with the same formulation successfully before. The only the difference between the old and present problem is the size. It's unbelivable. Anyway, I will try to check the code carefully again.

from sedumi.

Gurobi cannot solve SDPs. You've got something messed up in your YALMIP model, and would probably be better off posting a reproducible example on the YALMIP Google groups support forum. My uess you are running out of memory, if it really is really large. You still haven't showed us what the SeDuMi log is when you run with debug set to 1 in sdpsettings though

from sedumi.

I have set the 'debug' to 1 in sdpsettings, but getting the same error message as shown in my first comment. I can't understand it.

from sedumi.

Do something like this then

`optimize(x-y>=0,x+y,sdpsettings('solver','sedumi','savedebug',1));

load sedumidebug

sedumi(F_struc(:,2:end),F_struc(:,1),c,K)
`

from sedumi.

It shows:
The problem is primal infeasible, there is no x such that Ax=b.
misuse function pretransfo

from sedumi.

On my model or your model? I just gave you an example of something that crashed (a trivially unbounded (primal infeasible) model. If you get the same on your model, well then you have the answer.

from sedumi.

It is pn your model.

from sedumi.

But the error message is not the same as mine.

from sedumi.

Sorry, I tried the 'savedebug' option in my model and got the same error message as yours.

from sedumi.

I think we find the error now.

from sedumi.

You are of course supposed to change the code to use your model, and then load the debug model and call sedumi manually and then show what is displayed

from sedumi.

Hello, johan. I have solved the problem. It is due to a parameter which is set to inf by mistake. After changing it to zero, the model works. Thanks for your help. You are so kind a man.

from sedumi.

@Yuqi-Wang Can you explain what it is that was set to infinity? I am facing the same problem right now but can't figure out what's wrong.

from sedumi.

@ArjunM21, on the YALMIP discussion you are not in any sense talking about a problem related to this. You have shown that SeDuMi simply crashes immediately due to an incorrect call to symbchol, and inconsistency in the number of arguments, indicating you have multiple versions installed, or some weird edited version.

from sedumi.