Junan Lin

Junan Lin

PhD student @ University of Oxford

About Me

Hi! I am Junan, a PhD student under the supervision of Prof. Luca Furieri and Prof. Paul Goulart in the University of Oxford. I previously finished my master study in Robotics, Systems and Control program in ETH Zürich in 2025, and my bachelor study in Beihang University in 2022.

My research lies at the intersection of machine learning, optimization, and control, with a current focus on learning-to-optimize methods for structured convex optimization solvers. My recent preprint, Learning Over-Relaxation Policies for ADMM with Convergence Guarantees, studies how to learn adaptive relaxation policies for ADMM to accelerate repeated quadratic-program solves while preserving convergence guarantees and avoiding the refactorization cost of penalty updates.

In my leisure time, I like travelling by driving (have already driven more than 15,000 km), swimming and singing.

Research Interests

  • Learning-to-Optimize: learned optimizer components, adaptive algorithm parameters, generalization and convergence guarantees.
  • Convex Optimization and First-Order Methods: ADMM, operator splitting, OSQP, quadratic programming, monotone operator methods.
  • Control and Decision Making: model predictive control, differentiable optimization, learning-based control.
  • Optimization for Machine Learning: nonconvex optimization, large-scale ML training, and optimizer design for modern ML systems.

News

  • [Apr. 2026] Our preprint Learning Over-Relaxation Policies for ADMM with Convergence Guarantees is now available on arXiv.
  • [Oct. 2025] I am now a PhD student in the University of Oxford!
  • [May. 2025] I finish my master thesis titled “Data-Driven Terminal Cost Approximation for Short-Horizon Linear Quadratic MPC” in IfA in ETH Zürich, and obtain my master’s degree! A big thank you to Dr. Efe Balta, Riccardo Zuliani and Prof. John Lygeros for your generous guidance along my project!
  • [Aug. 2024] My internship paper “Enhancing Port Automation: A Novel Object Detection Pipeline for Container Ship Bays” is accepted by 2024 IEEE Sensors Conference! Thanks for those who helped and guided me through the way!