Skip to Main content Skip to Navigation
Conference papers

A new regret analysis for Adam-type algorithms

Abstract : In this paper, we focus on a theory-practice gap for Adam and its variants (AMSgrad, AdamNC, etc.). In practice, these algorithms are used with a constant first-order moment parameter β 1 (typically between 0.9 and 0.99). In theory, regret guarantees for online convex optimization require a rapidly decaying β 1 → 0 schedule. We show that this is an artifact of the standard analysis, and we propose a novel framework that allows us to derive optimal, data-dependent regret bounds with a constant β 1 , without further assumptions. We also demonstrate the flexibility of our analysis on a wide range of different algorithms and settings.
Document type :
Conference papers
Complete list of metadatas
Contributor : Panayotis Mertikopoulos <>
Submitted on : Monday, December 7, 2020 - 1:58:53 PM
Last modification on : Wednesday, December 16, 2020 - 4:08:41 AM


Files produced by the author(s)


  • HAL Id : hal-03043697, version 1


Ahmet Alacaoglu, Yura Malitsky, Panayotis Mertikopoulos, Volkan Cevher. A new regret analysis for Adam-type algorithms. ICML '20: The 37th International Conference on Machine Learning, 2020, Vienna, Austria. ⟨hal-03043697⟩



Record views


Files downloads