Accelerating quadratic optimization with reinforcement learning

J. Ichnowski, P. Jain, B. Stellato, G. Banjac, M. Luo, J. Gonzalez, I. Stoica, F. Borrelli and K. Goldberg

@inproceedings{Ichnowski2021:NeurIPS,
  author = {J. Ichnowski and P. Jain and B. Stellato and G. Banjac and M. Luo and J. Gonzalez and I. Stoica and F. Borrelli and K. Goldberg},
  title = {Accelerating quadratic optimization with reinforcement learning},
  booktitle = {Advances in Neural Information Processing Systems (NeurIPS)},
  year = {2021}
}

First-order methods for quadratic optimization such as OSQP are widely used for large-scale machine learning and embedded optimal control, where many related problems must be rapidly solved. These methods face two persistent challenges: manual hyperparameter tuning and convergence time to high-accuracy solutions. To address these, we explore how Reinforcement Learning (RL) can learn a policy to tune parameters to accelerate convergence. In experiments with well-known QP benchmarks we find that our RL policy, RLQP, significantly outperforms state-of-the-art QP solvers by up to 3x. RLQP generalizes surprisingly well to previously unseen problems with varying dimension and structure from different applications, including the QPLIB, Netlib LP and Maros-Mészáros problems.