000226287 001__ 226287
000226287 005__ 20190902134714.0
000226287 037__ $$aCONF
000226287 245__ $$aFaster Coordinate Descent via Adaptive Importance Sampling
000226287 269__ $$a2017
000226287 260__ $$bPMLR$$c2017$$aUSA
000226287 336__ $$aConference Papers
000226287 520__ $$aCoordinate descent methods employ random partial updates of decision variables in order to solve huge-scale convex optimization problems. In this work, we introduce new adaptive rules for the random selection of their updates. By adaptive, we mean that our selection rules are based on the dual residual or the primal-dual gap estimates and can change at each iteration. We theoretically characterize the performance of our selection rules and demonstrate improvements over the state-of-the-art, and extend our theory and algorithms to general convex objectives. Numerical evidence with hinge-loss support vector machines and Lasso confirm that the practice follows the theory.
000226287 6531_ $$aml-ai
000226287 700__ $$aPerekrestenko, Dmytro
000226287 700__ $$0243957$$g199128$$aCevher, Volkan
000226287 700__ $$aJaggi, Martin$$g276449$$0250160
000226287 7112_ $$dApril 20-22, 2017$$cFort Lauderdale, Florida, USA$$a20th International Conference on Artificial Intelligence and Statistics (AISTATS) 2017
000226287 773__ $$j54$$tProceedings of the 20th International Conference on Artificial Intelligence and Statistics
000226287 8564_ $$zURL$$uhttp://proceedings.mlr.press/v54/perekrestenko17a.html
000226287 8564_ $$zn/a$$yn/a$$uhttps://infoscience.epfl.ch/record/226287/files/284.pdf$$s467788
000226287 8564_ $$zn/a$$uhttps://infoscience.epfl.ch/record/226287/files/284-supp.pdf$$s542364
000226287 8560_ $$fgosia.baltaian@epfl.ch
000226287 909C0 $$xU12179$$pLIONS$$0252306
000226287 909C0 $$pMLO$$0252581$$xU13319
000226287 909CO $$qGLOBAL_SET$$pconf$$pSTI$$pIC$$ooai:infoscience.tind.io:226287
000226287 917Z8 $$x231598
000226287 917Z8 $$x231598
000226287 917Z8 $$x252028
000226287 917Z8 $$x252028
000226287 917Z8 $$x276449
000226287 937__ $$aEPFL-CONF-226287
000226287 973__ $$rREVIEWED$$aEPFL
000226287 980__ $$aCONF