Open JackDevine opened 5 years ago
This is a problem with adaptive_rho
. If you disable it, OSQP converges without problems. I have recreated your script in python and opened an issue in the main repository about this: https://github.com/oxfordcontrol/osqp/issues/151
This seems like it can be closed in favor of the upstream issue. It isn't a bug in OSQP.jl.
Hi all,
Thanks for the hard work put into OSQP.
When using OSQP.jl from JuMP.jl to solve a quadratic programming problem, I find that OSQP does not satisfy the constraints that I require.
The first bit of code is just some code to make some random labeled data, where each data point is a 2D vector in $[-1,1]\times[-1,1]$. The points are separated by a straight line. You can ignore this code, you just need it to be able to run the example that I give
The next bit of code is the important bit:
Upon running this, I usually find that the minimum value of α is much less than zero, which is a violation of the constraint that I gave. If I use Ipopt and uncomment the commented line, then the constraint will be satisfied.