Open GeoffNN opened 4 years ago
This can be inspired by the trivial example https://github.com/openopt/chop/blob/master/examples/optim_dynamics.py
It would also be really nice to benchmark algorithms on nonconvex problems. Do the usual practical acceleration methods (backtracking line search) work here?
backtracking does work (in theory and practice) on non-convex problems
I meant does it work for accelerating convergence in practice on non-convex problems?
Yes
On Tue, Nov 24, 2020, 21:14 Geoffrey Negiar notifications@github.com wrote:
I meant does it work for accelerating convergence in practice on non-convex problems?
— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/openopt/copt/issues/68#issuecomment-733417817, or unsubscribe https://github.com/notifications/unsubscribe-auth/AACDZB4VGE3G7O74IZECN7DSRRSA3ANCNFSM4OCDSFLQ .
(as long as we're talking about deterministic algorithms, the top comment makes it confusing)
It would be nice to have an example to compare speed of convergence of SAGA/SVRG/SFW on problems attaining the same optimum.