ML-guided exploration and discovery

Data-driven exploration and discovery in science and engineering are challenging when dynamics are complex and data are limited. I develop methods that use probabilistic machine learning models to guide exploration based on uncertainty quantification.

  • Searching for simple/interpretable solutions in black-box optimization [1].
    • Simultaneously optimize objective and solution simplicity, by drawing connections between multi-objective Bayesian optimization and regularized acquisition functions.
    • Simple solutions provide interprebilty to the outcome and reduce cost of deployment and maintenance.

  • Learning to control robotic system with safety constraints [2].
    • Developed probabilistic safety certificates for control barrier functions based on uncertainty quantification.

[1] Sparse Bayesian Optimization. Paper
Sulin Liu* (equal contr.), Qing Feng*, David Eriksson*, Benjamin Letham, Eytan Bakshy.
International Conference on Artificial Intelligence and Statistics (AISTATS), 2023.

[2] ProBF : Probabilistic Safety Certificates with Barrier Functions. Paper | Code
Athindran Ramesh Kumar*, Sulin Liu* (equal contr., random order), Jaime F. Fisac, Ryan P. Adams, Peter J. Ramadge. Preprint, 2021.
short version at NeurIPS Safe and Robust Control of Uncertain Systems Workshop, 2021.