MiniMax L2O: A Pilot Study
TL; DR - We introduce the learning to optimize (L2O) methodology to the minimax problems for the first time.
Learning to Optimize (L2O) is a subset of machine learning (ML) that seeks to combine the theoretical guarantees of classic optimization methods with the exceptional performance of data-driven algorithms. This is my primary area of research and the underlying theme of numerous blog posts below. These posts present the core ideas of mathematical research, typically one for each published work. Feel free to browse, comment, and reach out if you wish to connect and discuss these materials in further detail. I am quite open to collaboration on new projects.
Posts by Year
2021
2020
Projecting to Manifolds via Unsupervised Learning
TL; DR - We present an algorithm for performing projections onto the low dimensional manifolds that efficiently represent true data.
Safeguarded Learned Convex Optimization
TL; DR - Safeguarded learning to optimize (L2O) algorithms can leverage the strength of machine learning tools while maintaining convergence guarantees.
2019
Async Inertial Method for Common Fixed Points
TL; DR - Baby version of ARock for common fixed points without using any probability.