Optimal Hessian/Jacobian-Free Nonconvex-PL Bilevel Optimization

Abstract

Bilevel optimization is widely applied in many machine learning tasks such as hyper-parameter learning, meta learning and reinforcement learning. Although many algorithms recently have been developed to solve the bilevel optimization problems, they generally rely on the (strongly) convex lower-level problems. More recently, some methods have been proposed to solve the nonconvex-PL bilevel optimization problems, where their upper-level problems are possibly nonconvex, and their lower-level problems are also possibly nonconvex while satisfying Polyak-Łojasiewicz (PL) condition. However, these methods still have a high convergence complexity or a high computation complexity such as requiring compute expensive Hessian/Jacobian matrices and its inverses. In the paper, thus, we propose an efficient Hessian/Jacobian-free method (i.e., HJFBiO) with the optimal convergence complexity to solve the nonconvex-PL bilevel problems. Theoretically, under some mild conditions, we prove that our HJFBiO method obtains an optimal convergence rate of $O(\frac{1}{T})$, where $T$ denotes the number of iterations, and has an optimal gradient complexity of $O(\epsilon^{-1})$ in finding an $\epsilon$-stationary solution. We conduct some numerical experiments on the bilevel PL game and hyper-representation learning task to demonstrate efficiency of our proposed method.

Cite

Text

Huang. "Optimal Hessian/Jacobian-Free Nonconvex-PL Bilevel Optimization." International Conference on Machine Learning, 2024.

Markdown

[Huang. "Optimal Hessian/Jacobian-Free Nonconvex-PL Bilevel Optimization." International Conference on Machine Learning, 2024.](https://mlanthology.org/icml/2024/huang2024icml-optimal/)

BibTeX

@inproceedings{huang2024icml-optimal,
  title     = {{Optimal Hessian/Jacobian-Free Nonconvex-PL Bilevel Optimization}},
  author    = {Huang, Feihu},
  booktitle = {International Conference on Machine Learning},
  year      = {2024},
  pages     = {19598-19621},
  volume    = {235},
  url       = {https://mlanthology.org/icml/2024/huang2024icml-optimal/}
}