Originally posted by: Mambo12345

Good evening,

it seems that [x,fval,exitflag,output] = cplexmiqp(H, f, A, b', Aeq, beq', [], [], [], [], [], ctype ,[],options);

gives me a wrong solution. When I turn the iteration on to see what it does I can see that it cuts of so many branches.
That are my only options:

options.Display = 'iter';
Cplex.Param.mip.cuts.bqp = 3;
Cplex.Param.emphasis.numerical = 1;

But even when I didnt set any option it gave me a wrong solution.
What could be the mistake ?

Best regards !

Hard to tell without knowing why you thing the solution is wrong. Is it infeasible? Is it sub-optimal?

Can you also show the log output? Even better, share the actual model?

Also note that the way you the bqp and numerical emphasis parameters will take no effect: You are trying (in a wrong way) to change parameters for the class API while you use the toolbox API. Can you please check the reference documentation of cplexoptimset() (like I asked you to do in the other thread)? It has an explicit example how to set parameters for the toolbox API:

To turn on the optimizer output and set a node limit of 400:

 options = cplexoptimset(
'Display', 
'on', 
'MaxNodes', 400);

To set a node limit of 400 and instruct CPLEX to use traditional branch and cut style search, you have two methods:

 opt = cplexoptimset(
'mip.limits.nodes',
'400',
'mip.strategy.search',
'1');

or

 opt = cplexoptimset(
'cplex'); 
 opt.mip.limits.nodes=400; 
 opt.mip.strategy.search=1;

If you want to write the model out to a file, then you need to set:

 opt.exportmodel = 
'myModel.lp';

So to set numerical emphasis you would need something like options.emphasis.numerical = 1;

This is also explained in detail in the documentation here.

Originally posted by: Mambo12345

Thank you for your help !
I found a little mistake in my implementation and thats why i got a wrong solution.

That was my mistake but I still have the problem that I think it is very slow, I postet my code in the other thread, maybe you can find what makes it so slow.
Or maybe it is not possible to make it faster.