Randomized Stochastic Gradient Descent Ascent
arXiv:2111.13162
Abstract
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.
References in corpus (10)
- Efficient Algorithms for Smooth Minimax Optimization
- Accelerated Zeroth-Order and First-Order Momentum Methods from Mini to Minimax Optimization
- Optimal Epoch Stochastic Gradient Descent Ascent Methods for Min-Max Optimization
- Learning with minibatch Wasserstein : asymptotic and gradient properties
- A Single-Loop Smoothed Gradient Descent-Ascent Algorithm for Nonconvex-Concave Min-Max Problems
- Single-Timescale Stochastic Nonconvex-Concave Optimization for Smooth Nonlinear TD Learning
- Proximal Gradient Descent-Ascent: Variable Convergence under KŁ Geometry
- Tighter Analysis of Alternating Stochastic Gradient Method for Stochastic Nested Problems
- Almost sure convergence rates for Stochastic Gradient Descent and Stochastic Heavy Ball
- On the Existence of Optimal Transport Gradient for Learning Generative Models