Bundle-level type methods uniformly optimal for smooth and nonsmooth convex optimization (Q2515032): Difference between revisions

The author considers is the convex programming problem (CP) \[ f^{\ast }:=\min_{x\in X}f\left( x\right) , \] where \(X\) is a convex compact set and \(f:X\rightarrow \mathbb{R}\) is a closed function. The author develops uniformly optimal first-order methods for (CP). By incorporating a multistep acceleration scheme into the well-known bundle-level method, he develops an accelerated bundle-level method, and shows that it can achieve the optimal complexity for solving a general class of black-box CP problems without requiring the input of any smoothness information. The author develops a more practical, restricted memory version of this method, namely the accelerated prox-level (APL) method. He investigate the generalization of this method, namely the USL method, for solving certain composite CP problems and an important class of saddle-point problems. Finally, the author shows the advantages of the APL and USL methods over some existing first-order methods for solving certain classes of semidefinite programming and stochastic programming problems.