Randomized Stochastic Gradient Descent Ascent - Archive ouverte HAL Accéder directement au contenu
Communication Dans Un Congrès Année : 2022

Randomized Stochastic Gradient Descent Ascent

Othmane Sebbouh
  • Fonction : Auteur
  • PersonId : 1050009
Marco Cuturi
  • Fonction : Auteur
  • PersonId : 1085872

Résumé

An increasing number of machine learning problems, such as robust or adversarial variants of existing algorithms, require minimizing a loss function that is itself defined as a maximum. Carrying a loop of stochastic gradient ascent (SGA) steps on the (inner) maximization problem, followed by an SGD step on the (outer) minimization, is known as Epoch Stochastic Gradient \textit{Descent Ascent} (ESGDA). While successful in practice, the theoretical analysis of ESGDA remains challenging, with no clear guidance on choices for the inner loop size nor on the interplay between inner/outer step sizes. We propose RSGDA (Randomized SGDA), a variant of ESGDA with stochastic loop size with a simpler theoretical analysis. RSGDA comes with the first (among SGDA algorithms) almost sure convergence rates when used on nonconvex min/strongly-concave max settings. RSGDA can be parameterized using optimal loop sizes that guarantee the best convergence rates known to hold for SGDA. We test RSGDA on toy and larger scale problems, using distributionally robust optimization and single-cell data matching using optimal transport as a testbed.

Dates et versions

hal-03869529 , version 1 (24-11-2022)

Identifiants

Citer

Othmane Sebbouh, Gabriel Peyré, Marco Cuturi. Randomized Stochastic Gradient Descent Ascent. 25th International Conference on Artificial Intelligence and Statistics, Mar 2022, Online, France. ⟨hal-03869529⟩
18 Consultations
0 Téléchargements

Altmetric

Partager

Gmail Facebook X LinkedIn More