Surrogate Losses for Online Learning of Stepsizes in Stochastic Non-Convex Optimization
Zhenxun Zhuang, Ashok Cutkosky, Francesco Orabona
Abstract
Stochastic Gradient Descent (SGD) has played a central role in machine learning. However, it requires a carefully hand-picked stepsize for fast convergence, which is notoriously tedious and time-consuming to tune. Over the last several years, a plethora of adaptive gradient-based algorithms have emerged to ameliorate this problem. In this paper, we propose new surrogate losses to cast the problem of learning the optimal stepsizes for the stochastic optimization of a non-convex smooth objective function onto an online convex optimization problem. This allows the use of no-regret online algorithms to compute optimal stepsizes on the fly. In turn, this results in a SGD algorithm with self-tuned stepsizes that guarantees convergence rates that are automatically adaptive to the level of noise.
BibTeX
@InProceedings{pmlr-v97-zhuang19a,
title = {Surrogate Losses for Online Learning of Stepsizes in Stochastic Non-Convex Optimization},
author = {Zhuang, Zhenxun and Cutkosky, Ashok and Orabona, Francesco},
booktitle = {Proceedings of the 36th International Conference on Machine Learning},
pages = {7664--7672},
year = {2019},
editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan},
volume = {97},
series = {Proceedings of Machine Learning Research},
month = {09--15 Jun},
publisher = {PMLR},
pdf = {http://proceedings.mlr.press/v97/zhuang19a/zhuang19a.pdf},
url = {https://proceedings.mlr.press/v97/zhuang19a.html},
abstract = {Stochastic Gradient Descent (SGD) has played a central role in machine learning. However, it requires a carefully hand-picked stepsize for fast convergence, which is notoriously tedious and time-consuming to tune. Over the last several years, a plethora of adaptive gradient-based algorithms have emerged to ameliorate this problem. In this paper, we propose new surrogate losses to cast the problem of learning the optimal stepsizes for the stochastic optimization of a non-convex smooth objective function onto an online convex optimization problem. This allows the use of no-regret online algorithms to compute optimal stepsizes on the fly. In turn, this results in a SGD algorithm with self-tuned stepsizes that guarantees convergence rates that are automatically adaptive to the level of noise.}
}