OCP-GN: A Scalable Second-order Optimizer for Stochastic Optimization
Computer Vision and Pattern Recognition
2026-05-12 v2 Optimization and Control
Abstract
This paper proposes a novel second-order optimization algorithm based on the Optimal Control Principle (OCP), applicable to large-scale optimization problems in neural network training. The algorithm has a computational complexity of O(d) and strong robustness. Extensive experiments on multiple benchmarks demonstrate the significant superiority of the proposed method.
Cite
@article{arxiv.2512.24552,
title = {OCP-GN: A Scalable Second-order Optimizer for Stochastic Optimization},
author = {Jindi Zhong and Congyaohui Yin and Zhaorong Zhang and Huanshui Zhang},
journal= {arXiv preprint arXiv:2512.24552},
year = {2026}
}