Differences between models implemented with cvxpy and pyomo

Hi all, I’m trying to reproduce the results in a python library called pyStoned using cvxpy (they use pyomo) and we’re getting different solutions using the same solver (MOSEK) and I would like to understand why. For example, the model I’m trying to compare is the Convex Nonparametric Least Squares (CNLS):

The experiment with code is available here

This forum deal with CVX, which runs under MATLAB. So you will need to seek assistance on Python tools elsewhere.

This is neither a forum about Pyomo nor CVXPY.

However if you are getting worryingly different solutions to a problem that should have a unique solution (we are not talking about small floating-point noise) with MOSEK then feel free to dump the task files or provide other reproducible code to the Mosek support email and we will take a look at it there.

1 Like

BTW I wrote you briefly in that github issue but I only glanced at the log outputs.

1 Like

Oh, sorry for the confusion