English

Natural Hypergradient Descent: Algorithm Design, Convergence Analysis, and Parallel Implementation

Machine Learning 2026-04-02 v2 Optimization and Control Machine Learning

Abstract

In this work, we propose Natural Hypergradient Descent (NHGD), a new method for solving bilevel optimization problems. To address the computational bottleneck in hypergradient estimation--namely, the need to compute or approximate Hessian inverse--we exploit the statistical structure of the inner optimization problem and use the empirical Fisher information matrix as an asymptotically consistent surrogate for the Hessian. This design enables a parallel optimize-and-approximate framework in which the Hessian-inverse approximation is updated synchronously with the stochastic inner optimization, reusing gradient information at negligible additional cost. Our main theoretical contribution establishes high-probability error bounds and sample complexity guarantees for NHGD that match those of state-of-the-art optimize-then-approximate methods, while significantly reducing computational time overhead. Empirical evaluations on representative bilevel learning tasks further demonstrate the practical advantages of NHGD, highlighting its scalability and effectiveness in large-scale machine learning settings.

Keywords

Cite

@article{arxiv.2602.10905,
  title  = {Natural Hypergradient Descent: Algorithm Design, Convergence Analysis, and Parallel Implementation},
  author = {Deyi Kong and Zaiwei Chen and Shuzhong Zhang and Shancong Mou},
  journal= {arXiv preprint arXiv:2602.10905},
  year   = {2026}
}