%0 Generic %1 pedregosa2018adaptive %A Pedregosa, Fabian %A Gidel, Gauthier %D 2018 %K fused-lasso optimization structured_sparsity total_variation trend_filter %T Adaptive Three Operator Splitting. %U http://arxiv.org/abs/1804.02339 %X We propose and analyze an adaptive step-size variant of the Davis-Yin three operator splitting. This method can solve optimization problems composed by a sum of a smooth term for which we have access to its gradient and an arbitrary number of potentially non-smooth terms for which we have access to their proximal operator. The proposed method sets the step-size based on local information of the objective --hence allowing for larger step-sizes--, only requires two extra function evaluations per iteration and does not depend on any step-size hyperparameter besides an initial estimate. We provide an iteration complexity analysis that matches the best known results for the non-adaptive variant: sublinear convergence for general convex functions and linear convergence under strong convexity of the smooth term and smoothness of one of the proximal terms. Finally, an empirical comparison with related methods on 6 different problems illustrates the computational advantage of the proposed method.

@misc{pedregosa2018adaptive, abstract = {We propose and analyze an adaptive step-size variant of the Davis-Yin three operator splitting. This method can solve optimization problems composed by a sum of a smooth term for which we have access to its gradient and an arbitrary number of potentially non-smooth terms for which we have access to their proximal operator. The proposed method sets the step-size based on local information of the objective --hence allowing for larger step-sizes--, only requires two extra function evaluations per iteration and does not depend on any step-size hyperparameter besides an initial estimate. We provide an iteration complexity analysis that matches the best known results for the non-adaptive variant: sublinear convergence for general convex functions and linear convergence under strong convexity of the smooth term and smoothness of one of the proximal terms. Finally, an empirical comparison with related methods on 6 different problems illustrates the computational advantage of the proposed method.}, added-at = {2018-12-07T10:30:33.000+0100}, author = {Pedregosa, Fabian and Gidel, Gauthier}, biburl = {https://www.bibsonomy.org/bibtex/257273e19af372158ebfefcde762035b1/jpvaldes}, description = {Adaptive Three Operator Splitting}, interhash = {245723100ea24704a2c3ad2708dd09ff}, intrahash = {57273e19af372158ebfefcde762035b1}, keywords = {fused-lasso optimization structured_sparsity total_variation trend_filter}, note = {cite arxiv:1804.02339}, timestamp = {2018-12-07T10:30:33.000+0100}, title = {Adaptive Three Operator Splitting.}, url = {http://arxiv.org/abs/1804.02339}, year = 2018 }