Stochastic Dual Coordinate Ascent with Alternating Direction Method of Multipliers

Taiji Suzuki
Proceedings of the 31st International Conference on Machine Learning, PMLR 32(1):736-744, 2014.

Abstract

We propose a new stochastic dual coordinate ascent technique that can be applied to a wide range of regularized learning problems. Our method is based on alternating direction method of multipliers (ADMM) to deal with complex regularization functions such as structured regularizations. Although the original ADMM is a batch method, the proposed method offers a stochastic update rule where each iteration requires only one or few sample observations. Moreover, our method can naturally afford mini-batch update and it gives speed up of convergence. We show that, under mild assumptions, our method converges exponentially. The numerical experiments show that our method actually performs efficiently.

Cite this Paper

BibTeX
@InProceedings{pmlr-v32-suzuki14, title = {Stochastic Dual Coordinate Ascent with Alternating Direction Method of Multipliers}, author = {Suzuki, Taiji}, booktitle = {Proceedings of the 31st International Conference on Machine Learning}, pages = {736--744}, year = {2014}, editor = {Xing, Eric P. and Jebara, Tony}, volume = {32}, number = {1}, series = {Proceedings of Machine Learning Research}, address = {Bejing, China}, month = {22--24 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v32/suzuki14.pdf}, url = {https://proceedings.mlr.press/v32/suzuki14.html}, abstract = {We propose a new stochastic dual coordinate ascent technique that can be applied to a wide range of regularized learning problems. Our method is based on alternating direction method of multipliers (ADMM) to deal with complex regularization functions such as structured regularizations. Although the original ADMM is a batch method, the proposed method offers a stochastic update rule where each iteration requires only one or few sample observations. Moreover, our method can naturally afford mini-batch update and it gives speed up of convergence. We show that, under mild assumptions, our method converges exponentially. The numerical experiments show that our method actually performs efficiently.} }
Endnote
%0 Conference Paper %T Stochastic Dual Coordinate Ascent with Alternating Direction Method of Multipliers %A Taiji Suzuki %B Proceedings of the 31st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2014 %E Eric P. Xing %E Tony Jebara %F pmlr-v32-suzuki14 %I PMLR %P 736--744 %U https://proceedings.mlr.press/v32/suzuki14.html %V 32 %N 1 %X We propose a new stochastic dual coordinate ascent technique that can be applied to a wide range of regularized learning problems. Our method is based on alternating direction method of multipliers (ADMM) to deal with complex regularization functions such as structured regularizations. Although the original ADMM is a batch method, the proposed method offers a stochastic update rule where each iteration requires only one or few sample observations. Moreover, our method can naturally afford mini-batch update and it gives speed up of convergence. We show that, under mild assumptions, our method converges exponentially. The numerical experiments show that our method actually performs efficiently.
RIS
TY - CPAPER TI - Stochastic Dual Coordinate Ascent with Alternating Direction Method of Multipliers AU - Taiji Suzuki BT - Proceedings of the 31st International Conference on Machine Learning DA - 2014/01/27 ED - Eric P. Xing ED - Tony Jebara ID - pmlr-v32-suzuki14 PB - PMLR DP - Proceedings of Machine Learning Research VL - 32 IS - 1 SP - 736 EP - 744 L1 - http://proceedings.mlr.press/v32/suzuki14.pdf UR - https://proceedings.mlr.press/v32/suzuki14.html AB - We propose a new stochastic dual coordinate ascent technique that can be applied to a wide range of regularized learning problems. Our method is based on alternating direction method of multipliers (ADMM) to deal with complex regularization functions such as structured regularizations. Although the original ADMM is a batch method, the proposed method offers a stochastic update rule where each iteration requires only one or few sample observations. Moreover, our method can naturally afford mini-batch update and it gives speed up of convergence. We show that, under mild assumptions, our method converges exponentially. The numerical experiments show that our method actually performs efficiently. ER -
APA
Suzuki, T.. (2014). Stochastic Dual Coordinate Ascent with Alternating Direction Method of Multipliers. Proceedings of the 31st International Conference on Machine Learning, in Proceedings of Machine Learning Research 32(1):736-744 Available from https://proceedings.mlr.press/v32/suzuki14.html.

Related Material