English

OPTAMI: Global Superlinear Convergence of High-order Methods

Optimization and Control 2024-10-15 v2

Abstract

Second-order methods for convex optimization outperform first-order methods in terms of theoretical iteration convergence, achieving rates up to O(k5)O(k^{-5}) for highly-smooth functions. However, their practical performance and applications are limited due to their multi-level structure and implementation complexity. In this paper, we present new results on high-order optimization methods, supported by their practical performance. First, we show that the basic high-order methods, such as the Cubic Regularized Newton Method, exhibit global superlinear convergence for μ\mu-strongly star-convex functions, a class that includes μ\mu-strongly convex functions and some non-convex functions. Theoretical convergence results are both inspired and supported by the practical performance of these methods. Secondly, we propose a practical version of the Nesterov Accelerated Tensor method, called NATA. It significantly outperforms the classical variant and other high-order acceleration techniques in practice. The convergence of NATA is also supported by theoretical results. Finally, we introduce an open-source computational library for high-order methods, called OPTAMI. This library includes various methods, acceleration techniques, and subproblem solvers, all implemented as PyTorch optimizers, thereby facilitating the practical application of high-order methods to a wide range of optimization problems. We hope this library will simplify research and practical comparison of methods beyond first-order.

Keywords

Cite

@article{arxiv.2410.04083,
  title  = {OPTAMI: Global Superlinear Convergence of High-order Methods},
  author = {Dmitry Kamzolov and Dmitry Pasechnyuk and Artem Agafonov and Alexander Gasnikov and Martin Takáč},
  journal= {arXiv preprint arXiv:2410.04083},
  year   = {2024}
}
R2 v1 2026-06-28T19:09:38.464Z