There are many similar problems ：MOSEK warning 53: A numerically large upper bound value 1.4e+08 is specified for constraint '' (347)
My problem is the yellow part. The scaling is refering to
If I delete the yellow part or change it , I can get a feasible solution. The yellow part generate the following warnings
MOSEK warns against large bounds because they can cause numerical issues and most likely does in this case.
[Maybe you should think about why those large bounds are needed. Maybe they can be removed or reduced.]
It seems Mosek computes a reasonable accurate solution despite the large bounds. For some reason
Cvx does not report the solution if Mosek does not declare it optimal. I would call it a bug but @mcg may call it a feature.
yalmip it supposed to work diffrently.
@Erling That’s interesting. The preceding post is an example of Mosek terminating with UNKNOWN but CVX status Inaccurate/Solved and providing the solution to the user. I don’t recall seeing this behavior before.
In each iteration, CVX all provides the status Inaccurate/Solved. I want to know whether this result can be regarded as the optimal value if convergence.
The optimizer says the solution is optimal if a certain criteria is satisfied.
Now this criteria is not given by god and hence the solution will always be an approximate answer.
So the answer to your questions is maybe.
In any case it is recommended to check whether the solution you produce makes sense. You are really the only one who can do that.