DREAM: Delay-Resistant Acceleration with Momentum
math.OC
Submitted: 2026-10-27
Updated: 2026-11-13
License: http://creativecommons.org/licenses/by/4.0/
The gist: With exact, delay-free gradients, Nesterov's accelerated gradient (NAG) achieves optimal sqrtκ dependence for L-smooth, μ-strongly convex objectives, where κ=L/μ…
Abstract
With exact, delay-free gradients, Nesterov's accelerated gradient (NAG) achieves optimal sqrtκ dependence for L-smooth, μ-strongly convex objectives, where κ=L/μ; heavy-ball (HB) achieves the same scaling on strongly convex quadratics. In asynchronous and distributed settings, gradients can be inexact, with delays varying across workers and iterations. We isolate staleness using a fixed delay τ at least1, replacing grad f(y k) with grad f(y k-τ) at iteration k. Arjevani, Shamir, and Srebro establish a worst-case lower bound of Ω(τ sqrtκ (1/epsilon)) iterations to reach objective error at most epsilon as epsilon to0. We show that for any fixed positive delay, neither NAG nor HB retains sqrtκ-acceleration on strongly convex quadratics when updating every iteration, for any stable constant step size and positive momentum. We introduce delay-resistant accelerated momentum (DREAM), which decouples gradient-query extrapolation from momentum and uses explicit delay-dependent parameters. DREAM attains a linear convergence rate 1-Θ(1/(τ sqrtκ)) for all smooth strongly convex objectives as κ to infinity, yielding complexity matching the delayed first-order lower bound up to logarithmic factors. Our proof combines characteristic-polynomial stability analysis on quadratics with an analytic Zames--Falb integral quadratic constraint certificate. We extend DREAM to smooth convex objectives, obtaining near-optimal iteration complexity up to logarithmic factors for sufficiently small target accuracy. Experiments on strongly convex quadratics and regularized logistic regression corroborate DREAM's accelerated condition-number dependence under delay, unlike NAG and HB, and suggest this scaling persists when parameters use a conservative bound on the true delay.
Sources
- Accelerating Asynchronous Algorithms for Convex Optimization by Momentum Compensation
- A Distributed Asynchronous Generalized Momentum Algorithm Without Delay Bounds
Related papers
- An Accelerated Proximal Gradient Method for Bures-Wasserstein Space
- FOMTools: Development and AI-Assisted Code Generation for First-Order Methods
- Lions and Muons: Optimization via Stochastic Frank-Wolfe under Heavy-Tailed Noise
- Dynamic Service Recommendation with Congestion-Dependent Joining: Near-Optimal and Constant-Factor Approximation Algorithms
- Integrating Learning and Optimization for Open Transport Capacity Pooling under Disruption
- Zeroth-Order Gradient Estimators under Decision-Dependent Noise