Distributed Box-Constrained Quadratic Optimization for Dual Linear SVM

Ching-Pei Lee, Dan Roth
Proceedings of the 32nd International Conference on Machine Learning, PMLR 37:987-996, 2015.

Abstract

Training machine learning models sometimes needs to be done on large amounts of data that exceed the capacity of a single machine, motivating recent works on developing algorithms that train in a distributed fashion. This paper proposes an efficient box-constrained quadratic optimization algorithm for distributedly training linear support vector machines (SVMs) with large data. Our key technical contribution is an analytical solution to the problem of computing the optimal step size at each iteration, using an efficient method that requires only O(1) communication cost to ensure fast convergence. With this optimal step size, our approach is superior to other methods by possessing global linear convergence, or, equivalently, O(\log(1/ε)) iteration complexity for an epsilon-accurate solution, for distributedly solving the non-strongly-convex linear SVM dual problem. Experiments also show that our method is significantly faster than state-of- the-art distributed linear SVM algorithms including DSVM-AVE, DisDCA and TRON.

Cite this Paper


BibTeX
@InProceedings{pmlr-v37-leea15, title = {Distributed Box-Constrained Quadratic Optimization for Dual Linear SVM}, author = {Lee, Ching-Pei and Roth, Dan}, booktitle = {Proceedings of the 32nd International Conference on Machine Learning}, pages = {987--996}, year = {2015}, editor = {Bach, Francis and Blei, David}, volume = {37}, series = {Proceedings of Machine Learning Research}, address = {Lille, France}, month = {07--09 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v37/leea15.pdf}, url = {https://proceedings.mlr.press/v37/leea15.html}, abstract = {Training machine learning models sometimes needs to be done on large amounts of data that exceed the capacity of a single machine, motivating recent works on developing algorithms that train in a distributed fashion. This paper proposes an efficient box-constrained quadratic optimization algorithm for distributedly training linear support vector machines (SVMs) with large data. Our key technical contribution is an analytical solution to the problem of computing the optimal step size at each iteration, using an efficient method that requires only O(1) communication cost to ensure fast convergence. With this optimal step size, our approach is superior to other methods by possessing global linear convergence, or, equivalently, O(\log(1/ε)) iteration complexity for an epsilon-accurate solution, for distributedly solving the non-strongly-convex linear SVM dual problem. Experiments also show that our method is significantly faster than state-of- the-art distributed linear SVM algorithms including DSVM-AVE, DisDCA and TRON.} }
Endnote
%0 Conference Paper %T Distributed Box-Constrained Quadratic Optimization for Dual Linear SVM %A Ching-Pei Lee %A Dan Roth %B Proceedings of the 32nd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2015 %E Francis Bach %E David Blei %F pmlr-v37-leea15 %I PMLR %P 987--996 %U https://proceedings.mlr.press/v37/leea15.html %V 37 %X Training machine learning models sometimes needs to be done on large amounts of data that exceed the capacity of a single machine, motivating recent works on developing algorithms that train in a distributed fashion. This paper proposes an efficient box-constrained quadratic optimization algorithm for distributedly training linear support vector machines (SVMs) with large data. Our key technical contribution is an analytical solution to the problem of computing the optimal step size at each iteration, using an efficient method that requires only O(1) communication cost to ensure fast convergence. With this optimal step size, our approach is superior to other methods by possessing global linear convergence, or, equivalently, O(\log(1/ε)) iteration complexity for an epsilon-accurate solution, for distributedly solving the non-strongly-convex linear SVM dual problem. Experiments also show that our method is significantly faster than state-of- the-art distributed linear SVM algorithms including DSVM-AVE, DisDCA and TRON.
RIS
TY - CPAPER TI - Distributed Box-Constrained Quadratic Optimization for Dual Linear SVM AU - Ching-Pei Lee AU - Dan Roth BT - Proceedings of the 32nd International Conference on Machine Learning DA - 2015/06/01 ED - Francis Bach ED - David Blei ID - pmlr-v37-leea15 PB - PMLR DP - Proceedings of Machine Learning Research VL - 37 SP - 987 EP - 996 L1 - http://proceedings.mlr.press/v37/leea15.pdf UR - https://proceedings.mlr.press/v37/leea15.html AB - Training machine learning models sometimes needs to be done on large amounts of data that exceed the capacity of a single machine, motivating recent works on developing algorithms that train in a distributed fashion. This paper proposes an efficient box-constrained quadratic optimization algorithm for distributedly training linear support vector machines (SVMs) with large data. Our key technical contribution is an analytical solution to the problem of computing the optimal step size at each iteration, using an efficient method that requires only O(1) communication cost to ensure fast convergence. With this optimal step size, our approach is superior to other methods by possessing global linear convergence, or, equivalently, O(\log(1/ε)) iteration complexity for an epsilon-accurate solution, for distributedly solving the non-strongly-convex linear SVM dual problem. Experiments also show that our method is significantly faster than state-of- the-art distributed linear SVM algorithms including DSVM-AVE, DisDCA and TRON. ER -
APA
Lee, C. & Roth, D.. (2015). Distributed Box-Constrained Quadratic Optimization for Dual Linear SVM. Proceedings of the 32nd International Conference on Machine Learning, in Proceedings of Machine Learning Research 37:987-996 Available from https://proceedings.mlr.press/v37/leea15.html.

Related Material