The RATS Software Forum

Hey,

i have some problems using the lqprog or more procisly undertanding the error message and its reason: ## LPQ5. Active set method does not converge.
When i activate the trace comand, i got the feeling it does converge. I have created a file with four examples. Example 1,3 and 4 are leading to the error message. 2 does not. I have made coment in the file according the difference between the examples. Maybe someone is able to explain me, why the examples are leading to the errors.

Best Regards

The first one is easy---the 3rd and 6th variables aren't identified.

In general, all of your Q matrices aren't positive semi-definite; all have a slightly negative eigenvalue, but not a trivially small one. I don't know if you've lost too much precision in copying the information from another source, or whether that's a design flaw generally. Yes, it does look like they're converged, but the Lagrange multipliers are likely not settling down and that's required for it to be sure that it's converged.

Thanks Tom for this answer. Its helping a lot.
That variable 3 and 6 are not identified in the first example was desired, so i already thought that this might a problem. Since i got the other examples too, i was still wondering what might cause the problems. Thanks to your answer i know it now and can check for it.