not converge but solution status optimal

[Yixiao Yang]

Hello, I'm trying to use Mosek and Yalmip to solve a conic optimization problem, the solution status demonstrated optimal, but some elements didn't converge to zero. 

Increasing iteration steps is not useful, and I try to adjust MSK_DPAR_INTPNT_TOL_PFEAS, MSK_DPAR_INTPNT_TOL_DFEAS and MSK_IPAR_PRESOLVE_USE, the result has no modification.

Then I try to adjust MSK_DPAR_INTPNT_TOL_NEAR_REL, but caused NaN in result. Can you help me to find out where is the problem?

[Michal Adamaszek]

You would have to attach a task file or something else that demonstrates the issue.

You can never expect that the solution is exact no matter how tight tolerances you set, even if you know in advance that the optimal solution has x=0 then something like x=1e-8 is to be expected. See for example https://docs.mosek.com/latest/faq/faq.html#why-does-the-solution-slightly-violate-constraints

[Yixiao Yang]

sorry for not attaching files, here are mosek_log for the third optimization problem(marked online problem in code) starting from line 193 in code, which shows issues.

and the entire code file comes from a paper, and it is explicitly stated that the MOSEK solver is used to solve the optimization problem.