Comments

1. By Kim, on May 08, 2012, at 07:26 AM
   Thank you!
   
2. By Mohammad Karimadini, on May 08, 2012, at 08:59 AM
   Many thanks!
   
3. By Wuhua Hu, on May 08, 2012, at 09:39 AM
   Hi Johan. Do you know how to define and add user cuts in YALMIP (and Matlab) for the cplex solver to solve mixed-integer problems by branch and cut?
   
4. By johan, on May 08, 2012, at 11:38 AM
   Do you mean dynamically? There is no support for that through YALMIP.
   
5. By leo, on May 09, 2012, at 08:25 AM
   Many thanks!
   
6. By Mauricio, on May 09, 2012, at 07:54 PM
   Hi Johan,
   when I run the following test model using BONMIN, I get an error:
   
   x1 = sdpvar(50,1);
   x2 = intvar(50,1);
   x3 = intvar(50,1);
   
   objective = sum(2*x1+x2.*x3+x3.^2);
   
   F = [];
   F = F + set(x1'*(4*x1-4*x2+4*x3-20)+x2'*(2*x2-2*x3+9)+x3'*(2*x3-13)+24 > 0);
   
   F = F + set(500-x1'*x2 > 0);
   F = F + set(0 <= x1 <= 100);
   F = F + set(0 <= x2 <= 100);
   F = F + set(0 <= x3 <= 100);
   
   ops = sdpsettings('solver','bonmin');
   
   status = solvesdp(F,objective,ops);
   
   Unknown problem in solver (try using 'debug'-flag in sdpsettings) (Error using bonmin *** Error using Bonmin Matlab interface: *** You have specified a nonexistent BONMIN/IPOPT option ("req_solver"). )
   
   Regards,
   
7. By johan, on May 09, 2012, at 08:22 PM
   There seem to be a bug in the 1.55 release of optilab.
   
   This fixes the issue
   ops.bonmin.ipopt = rmfield(ops.bonmin.ipopt,'req_solver')
   
   Note that the problem is nonconvex so bonmin is not guaranteed to find a solution. You can just as well use 'bnb' which seems to perform better here, and if you want to obtain a certified global solution, you can use 'bmibnb'
   
   (changing lower bound on x to negative number, otherwise the problem is trivial for both 'bnb' and 'bmibnb')
   
8. By Mauricio, on May 10, 2012, at 02:05 AM
   Thanks a lot!
   
9. By johan, on May 10, 2012, at 08:18 AM
   Jonathan will release an updated version of OPTI Toolbox soon to fix these issue.
   
   I noticed some other issues related to OPTI Toolbox (with clp, cbc amd ooqp) so I will release a patched version today or tomorrow.
   
10. By johan, on May 11, 2012, at 12:23 PM
    A new version of OPTI Toolbox as been released.
    
11. By Jiang Yilang, on June 06, 2012, at 09:58 AM
    Hi Johan,
    when I run the following codes, I get an error:
    
    A_ = (Plant_Balancmr.C)*Plant_Balancmr.A*inv(Plant_Balancmr.C)
    Ad_ = (Plant_Balancmr.C)*Plant_Balancmr.B
    P = sdpvar(7, 7);
    Q = sdpvar(7, 7);
    V = sdpvar(7, 7);
    W = sdpvar(7, 7, 'full');
    d = sdpvar(1, 1);
    O(7, 7) = 0;
    Aug = [(A_+Ad_)'*P+P*(A_+Ad_)+W'*Ad_+Ad_'*W+Q -W'*Ad_ A_'*Ad_'*V d*(W+P)
    -Ad_'*W -Q Ad_'*Ad_'*V O
    V*Ad_*A_ V*Ad_*Ad_ -V O
    d*(W'+P) O O -V];
    Oaug(28, 28) = 0;
    F = [Aug<0];
    solvesdp(F, -d)
    y = max(real(eig(double(Aug))));
    %------------------------------
    %------------------------------
    the information is as follows:
    info: 'No suitable solver'
    
    How can I tackle this problem? Thank you.
    Best WIshes!
    Yours
    Jiang Yilang
    
12. By johan, on June 07, 2012, at 08:15 AM
    You have products between decision variables (d and W+P), hence, it is not problem with linear matrix inequalities (guessing you have sedumi or similar linear sdp solver installed)
    
    Your problem can easily be attacked using a bisection approach though
    http://users.isy.liu.se/johanl/yalmip/pmwiki.php?n=Examples.DecayRate