ASAGA: Asynchronous Parallel SAGA

Abstract

We describe ASAGA, an asynchronous parallel version of the incremental gradient algorithm SAGA that enjoys fast linear convergence rates. Through a novel perspective, we revisit and clarify a subtle but important technical issue present in a large fraction of the recent convergence rate proofs for asynchronous parallel optimization algorithms, and propose a simplification of the recently introduced "perturbed iterate" framework that resolves it. We thereby prove that ASAGA can obtain a theoretical linear speedup on multi-core systems even without sparsity assumptions. We present results of an implementation on a 40-core architecture illustrating the practical speedup as well as the hardware overhead.

Appears in: Proceedings of the 20th International Conference on Artificial Intelligence and Statistics (AISTATS 2017), 37 pages

Cited by in corpus (9)

• Federated Learning with Non-IID Data
• Federated Optimization: Distributed Machine Learning for On-Device Intelligence
• Revisiting Distributed Synchronous SGD
• Stochastic, Distributed and Federated Optimization for Machine Learning
• Asynchronous Accelerated Proximal Stochastic Gradient for Strongly Convex Distributed Finite Sums
• Pufferfish: Communication-efficient Models At No Extra Cost
• DS-MLR: Exploiting Double Separability for Scaling up Distributed Multinomial Logistic Regression
• Block stochastic gradient descent for large-scale tomographic reconstruction in a parallel network
• POLO: a POLicy-based Optimization library