This PR features updates to mostly the oracle solver for the sparse case. We did not previously create X_reduced like we do for the dense case, but with this PR we do this for the sparse case too.
It also includes an addition to the hybrid solver, by which we now cache the squared col norms of X also in the sparse case. I don't think this makes much difference, however. Maybe it's not even worth it since we need to store p floats?
This PR features updates to mostly the oracle solver for the sparse case. We did not previously create X_reduced like we do for the dense case, but with this PR we do this for the sparse case too.
It also includes an addition to the hybrid solver, by which we now cache the squared col norms of X also in the sparse case. I don't think this makes much difference, however. Maybe it's not even worth it since we need to store p floats?