Online Convex Optimization in the Random Order Model

Dan Garber, Gal Korcia, Kfir Levy
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:3387-3396, 2020.

Abstract

Online Convex Optimization (OCO) is a powerful framework for sequential prediction, portraying the natural uncertainty inherent in data-streams as though the data were generated by an almost omniscient adversary. However, this view, which is often too pessimistic for real-world data, comes with a price. The complexity of solving many important online tasks in this adversarial framework becomes much worse than that of their offline and even stochastic counterparts. In this work we consider a natural random-order version of the OCO model, in which the adversary can choose the set of loss functions, but does not get to choose the order in which they are supplied to the learner; Instead, they are observed in uniformly random order. Focusing on two important families of online tasks, one in which the cumulative loss function is strongly convex (though individual loss functions may not even be convex), and the other being online $k$-PCA, we show that under standard well-conditioned-data assumptions, standard online gradient descent (OGD) methods become much more efficient in the random-order model. In particular, for the first group of tasks OGD guarantees poly-logarithmic regret. In the case of online $k$-PCA, OGD guarantees sublinear regret using only a rank-$k$ SVD on each iteration and memory linear in the size of the solution.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-garber20a, title = {Online Convex Optimization in the Random Order Model}, author = {Garber, Dan and Korcia, Gal and Levy, Kfir}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {3387--3396}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/garber20a/garber20a.pdf}, url = {https://proceedings.mlr.press/v119/garber20a.html}, abstract = {Online Convex Optimization (OCO) is a powerful framework for sequential prediction, portraying the natural uncertainty inherent in data-streams as though the data were generated by an almost omniscient adversary. However, this view, which is often too pessimistic for real-world data, comes with a price. The complexity of solving many important online tasks in this adversarial framework becomes much worse than that of their offline and even stochastic counterparts. In this work we consider a natural random-order version of the OCO model, in which the adversary can choose the set of loss functions, but does not get to choose the order in which they are supplied to the learner; Instead, they are observed in uniformly random order. Focusing on two important families of online tasks, one in which the cumulative loss function is strongly convex (though individual loss functions may not even be convex), and the other being online $k$-PCA, we show that under standard well-conditioned-data assumptions, standard online gradient descent (OGD) methods become much more efficient in the random-order model. In particular, for the first group of tasks OGD guarantees poly-logarithmic regret. In the case of online $k$-PCA, OGD guarantees sublinear regret using only a rank-$k$ SVD on each iteration and memory linear in the size of the solution.} }
Endnote
%0 Conference Paper %T Online Convex Optimization in the Random Order Model %A Dan Garber %A Gal Korcia %A Kfir Levy %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-garber20a %I PMLR %P 3387--3396 %U https://proceedings.mlr.press/v119/garber20a.html %V 119 %X Online Convex Optimization (OCO) is a powerful framework for sequential prediction, portraying the natural uncertainty inherent in data-streams as though the data were generated by an almost omniscient adversary. However, this view, which is often too pessimistic for real-world data, comes with a price. The complexity of solving many important online tasks in this adversarial framework becomes much worse than that of their offline and even stochastic counterparts. In this work we consider a natural random-order version of the OCO model, in which the adversary can choose the set of loss functions, but does not get to choose the order in which they are supplied to the learner; Instead, they are observed in uniformly random order. Focusing on two important families of online tasks, one in which the cumulative loss function is strongly convex (though individual loss functions may not even be convex), and the other being online $k$-PCA, we show that under standard well-conditioned-data assumptions, standard online gradient descent (OGD) methods become much more efficient in the random-order model. In particular, for the first group of tasks OGD guarantees poly-logarithmic regret. In the case of online $k$-PCA, OGD guarantees sublinear regret using only a rank-$k$ SVD on each iteration and memory linear in the size of the solution.
APA
Garber, D., Korcia, G. & Levy, K.. (2020). Online Convex Optimization in the Random Order Model. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:3387-3396 Available from https://proceedings.mlr.press/v119/garber20a.html.

Related Material