Supervised learning & Data-driven methods

The proliferation of high-resolution datasets in turbulence due to advances in computing and algorithms over the past decade allows us to use data science as a viable tool to study turbulence. Machine learning-driven models have already attained spectacular success in fields such as language translation, speech and face recognition, bioinformatics, and advertising.

One of our research topics is to adapt these tools and apply them to persistent problems in turbulence research such as turbulence modeling and drag reduction. 

Reinforcement learning

Reinforcement learning identifies optimal strategies for agents that perform actions, contingent on their information about the environment, and measure their performance via scalar reward functions. In this semi-supervised learning framework with foundations in dynamic programming, training is not performed on a database of reference data but is performed by integrating in time the model and its consequences to the flow field.

Applications for reinforcement learning in turbulence are diverse, such as turbulence modeling and flow control. Our goal is to utilize reinforcement learning in various applications involving turbulent flows.