kul-optec
@qp-message branch- ForBES message
ForBES gives messgae-"reached optimum" with adaptive nama and panoc on maros-meszaros- but problem not solved- for example choose two sets DTOC3 & DUALC2 with p_test_maros_meszaros.m
Are those problems strongly convex? If that is not the case, unfortunately NAMA will have a hard time at solving them.
On Fri, 18 Jan 2019 at 10:35, Manish [email protected] wrote:
ForBES gives messgae-"reached optimum" with adaptive nama and panoc on maros-meszaros- but problem not solved- for example choose two sets DTOC3 & DUALC2 with p_test_maros_meszaros.m
— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kul-forbes/ForBES/issues/6, or mute the thread https://github.com/notifications/unsubscribe-auth/AAafK51p6w9-4GH4acsWpoYpqt4lInKEks5vEZVrgaJpZM4aHi1M .
Yes, they are made strongly convex using regularization to Hessian. Could this message be maintained if the problem is not actually solved in ForBES? Actually the same type of data-sets fails line-search in a very naive implementation of NAMA.