Lower Bounds for Non-Convex Stochastic Optimization — mentioned by LolWolf

Thu, 11 Jun 2020 17:52:19 GMT

Depending on the level you're looking for, I'm always a huge fan of Rockafellar [0] for the more mathematical expositions (they're lovely, clear, and very well thought through). Even in this case, though, subgradients may not exist (they're only guaranteed to exist for convex functions and any nonconvex function has, almost by definition, at least one point for which there exists no subgradient), in which case one talks about sub-derivatives instead. These always exist whenever a function is semicontinuous (such as the "if" statement example given in the GGP comment), even if it is not convex…

Paper and discussion history. A mention is not an endorsement.

Lower Bounds for Non-Convex Stochastic Optimization · HackerPapers discussions

Lower Bounds for Non-Convex Stochastic Optimization — mentioned by LolWolf