yalmiptest('mosek') returns ' Reference to non-existent field 'sol' '

SK

unread,

Feb 15, 2015, 6:03:35 PM2/15/15

to yal...@googlegroups.com

Hello,

I am trying to make yalmip work with MOSEK.

First, I added yalmip to the path in Matlab as follows:

addpath(genpath('/home/<user_id>/Optimisation/yalmip'))

Then, similarly for MOSEK:

addpath(genpath('/home/<user_id>/Optimisation/mosek'))

When I run the test:

yalmiptest('mosek')

it finds the solver:

+++++++++++++++++++++++++++++++++++++++++++++++

| Searching for installed solvers |

+++++++++++++++++++++++++++++++++++++++++++++++

| Solver| Version/module| Status|

+++++++++++++++++++++++++++++++++++++++++++++++

| MOSEK| LP/QP| found|

| MOSEK| SOCP| found|

| MOSEK| SDP| found|

| MOSEK| GEOMETRIC| found|

| LINPROG| | found|

...............................................

but it cannot deal with any of the tests, i.e. for all the tests (except 'Core functionalities' and 'Bilinear SDP'), I get the following:

Solution: N/A

Solver message: Unknown problem in solver (try using 'debug'-flag in sdpsettings) (Reference to non-existent field 'sol'.)

What is causing this it and how could I make yalmip work with MOSEK?

My set up is as follows:

OS: ubuntu 14.04 LTS

Matlab: R2014b (8.4.0.150421)

Yalmip: R20150204

MOSEK: 7.1.0.20

Johan Löfberg

unread,

Feb 16, 2015, 1:57:25 AM2/16/15

to yal...@googlegroups.com

Try

sdpvar x

optimize(x>=0,x,sdpsettings('solver','mosek','debug',1))

and let me know where it crashes, i.e., the crash report

SK

unread,

Feb 16, 2015, 6:50:59 AM2/16/15

to yal...@googlegroups.com

Apparently, the problem was caused by the MOSEK license file which was put in a wrong place.

After fixing this, everything runs as expected:

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

| Test| Solution| Solver message|

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

| Core functionalities| N/A| Successfully solved (YALMIP)|

| LP| Correct| Successfully solved (MOSEK)|

| QP| Correct| Successfully solved (MOSEK)|

| SOCP| Correct| Successfully solved (MOSEK)|

| SDP| Correct| Successfully solved (MOSEK)|

| MAXDET| Correct| Successfully solved (MOSEK)|

| Infeasible LP| N/A| Infeasible problem (MOSEK)|

| Infeasible QP| N/A| Infeasible problem (MOSEK)|

| Infeasible SDP| N/A| Unbounded objective function (MOSEK)|

| Moment relaxation| Correct| Successfully solved (MOSEK)|

| Sum-of-squares| Correct| Successfully solved (MOSEK)|

| Bilinear SDP| N/A| Solver not applicable (mosek)|

+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++

Thank you for help.

Alexey Bodrov

unread,

May 26, 2018, 10:47:38 AM5/26/18

to YALMIP

Dear Johan,

I have faced the similar problem with yalmip, but MPT solver:

Reference to non-existent field 'requested_variables'.

Error in yalmip2mpt (line 19)

if isempty(interfacedata.requested_variables)

Error in callmpt3 (line 10)

Matrices = yalmip2mpt(interfacedata);

Error in optimizer/subsref (line 326)

eval(['output = ' self.model.solver.call '(self.model);']);

The solver is installed and test examples from the MPT website are working correctly:

+++++++++++++++++++++++++++++++++++++++++++++++

| Searching for installed solvers |

+++++++++++++++++++++++++++++++++++++++++++++++

| Solver| Version/module| Status|

+++++++++++++++++++++++++++++++++++++++++++++++

| MPT| 3| found|

What could be a problem? Other solvers ('quadprog' and 'ipopt') are working correctly.

Thank you!

Johan Löfberg

unread,

May 26, 2018, 10:53:55 AM5/26/18

to YALMIP

impossible to answer without reproducible example

Alexey Bodrov

unread,

May 26, 2018, 11:03:28 AM5/26/18

to YALMIP

I have tried to use your example:

sdpvar x

optimize(x>=0,x,sdpsettings('solver','mpt','debug',1))

However, it doesn't work. Again, both "quadprog" and "ipopt" are working fine.

Error using solvesdp (line 278)

You must specify parametric variables.

Error in optimize (line 31)

[varargout{1:nargout}] = solvesdp(varargin{:});

Johan Löfberg

unread,

May 26, 2018, 11:34:35 AM5/26/18

to YALMIP

MPT is not an LP solver, it is a multiparametric solver, so that test-case makes no sense

Reply all

Reply to author

Forward