We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo ScienceWISE logo

Statistics > Machine Learning

Title: Two-Tailed Averaging: Anytime Adaptive Once-in-a-while Optimal Iterate Averaging for Stochastic Optimization

Authors: Gábor Melis
Abstract: Tail averaging improves on Polyak averaging's non-asymptotic behaviour by excluding a number of leading iterates of stochastic optimization from its calculations. In practice, with a finite number of optimization steps and a learning rate that cannot be annealed to zero, tail averaging can get much closer to a local minimum point of the training loss than either the individual iterates or the Polyak average. However, the number of leading iterates to ignore is an important hyperparameter, and starting averaging too early or too late leads to inefficient use of resources or suboptimal solutions. Setting this hyperparameter to improve generalization is even more difficult, especially in the presence of other hyperparameters and overfitting. Furthermore, before averaging starts, the loss is only weakly informative of the final performance, which makes early stopping unreliable. To alleviate these problems, we propose an anytime variant of tail averaging, that has no hyperparameters and approximates the optimal tail at all optimization steps. Our algorithm is based on two running averages with adaptive lengths bounded in terms of the optimal tail length, one of which achieves approximate optimality with some regularity. Requiring only the additional storage for two sets of weights and periodic evaluation of the loss, the proposed two-tailed averaging algorithm is a practical and widely applicable method for improving stochastic optimization.
Subjects: Machine Learning (stat.ML); Computation and Language (cs.CL); Machine Learning (cs.LG)
Cite as: arXiv:2209.12581 [stat.ML]
  (or arXiv:2209.12581v1 [stat.ML] for this version)

Submission history

From: Gábor Melis [view email]
[v1] Mon, 26 Sep 2022 10:46:37 GMT (370kb,D)

Link back to: arXiv, form interface, contact.