01
Continuous-time reinforcement learning
Learning control policies for continuous dynamics from discrete-time observations.
Applied Mathematics · Control · Machine Learning
Martin Hernandez
Postdoctoral Scholar · Department of Statistics & Data Science · UCLA
I am an applied mathematician working at the interface of control theory, reinforcement learning, numerical analysis, and machine learning. My current research focuses on continuous-time reinforcement learning and mean-field control, with broader interests in randomized numerical methods, partial differential equations, and the mathematical foundations of learning.
Research
Mathematical methods for learning and control in continuous time.
My work combines analysis, numerical methods, and learning-based approaches, with an emphasis on mathematically structured models and algorithms.
02
Mean-field control and reinforcement learning
Control and learning for large interacting populations and mean-field systems.
03
Randomized numerical methods
Random batch and randomized decomposition methods for dynamical systems and PDEs.
04
Neural network approximation
Approximation theory, expressivity, and mathematical questions arising in deep learning.