OCLGNEFeb 2, 2022

Tight Convergence Rate Bounds for Optimization Under Power Law Spectral Conditions

arXiv:2202.00992v311 citations
AI Analysis

This work addresses the performance analysis of gradient-based optimization for researchers and practitioners, offering incremental improvements in theoretical bounds.

The paper tackles the problem of deriving tighter convergence rate bounds for optimization algorithms under power law spectral conditions, demonstrating unified optimal acceleration strategies and providing first proofs of tight lower bounds for Steepest Descent and Conjugate Gradients.

Performance of optimization on quadratic problems sensitively depends on the low-lying part of the spectrum. For large (effectively infinite-dimensional) problems, this part of the spectrum can often be naturally represented or approximated by power law distributions, resulting in power law convergence rates for iterative solutions of these problems by gradient-based algorithms. In this paper, we propose a new spectral condition providing tighter upper bounds for problems with power law optimization trajectories. We use this condition to build a complete picture of upper and lower bounds for a wide range of optimization algorithms -- Gradient Descent, Steepest Descent, Heavy Ball, and Conjugate Gradients -- with an emphasis on the underlying schedules of learning rate and momentum. In particular, we demonstrate how an optimally accelerated method, its schedule, and convergence upper bound can be obtained in a unified manner for a given shape of the spectrum. Also, we provide first proofs of tight lower bounds for convergence rates of Steepest Descent and Conjugate Gradients under spectral power laws with general exponents. Our experiments show that the obtained convergence bounds and acceleration strategies are not only relevant for exactly quadratic optimization problems, but also fairly accurate when applied to the training of neural networks.

Foundations

The foundational work for this paper's niche, ranked by how specifically the neighbourhood builds on it — not by global fame.

Your Notes