Hi guys and thank you for opensourcing this lovely admm solver.

Can you perhaps shed some light on the heuristic used to guess active constraints?

Also, if this guess is incorrect (i.e. misses a binding constraint), is it then correct to assume that the iterative refinement step will not converge towards the same solution as a problem where all active constraints are included?

If so, can one somehow detect and incorporate the missing constraint?

Thanks,

Marcus