Papers

Reinforcement Learning of Active Flow Control in Partially Observable Environments
M. Weissenbacher, A. Borovykh and G. Rigas
Flow Turbulence and Combustion, 2024
In-progress version presented at ERCOFTAC ML4FLUID workshop, 2024
Decay and non-decay for the massless Vlasov equation on subextremal and extremal Reissner-Nordström black holes
M. Weissenbacher
Archive for Rational Mechanics, 2024

Conferences

Model Based Reinforcement Learning for chaotic flow control
P. Gupta, M. Weissenbacher, G. Rigas
DFD annual meeting of the American Physical Society, 2024
Reinforcement learning for turbulent drag reduction of realistic road vehicles with dynamic flaps
J. Zhang, I. Fumarola, M. Weissenbacher, X. Jiang, G. Rigas
DFD annual meeting of the American Physical Society, 2024
Drag reduction for heavy road vehicles with rear flaps
X. Jiang, J. Zhang, M. Weissenbacher, I. Fumarola and G. Rigas
European Drag Reduction and Flow Control Meeting, 2024
Static and Dynamic Control for Wind Farm Wake Steering using LES
A. Mole, M. Weissenbacher, S. Laizet
1st European Fluid Dynamics Conference, 2024
CHAROT: Robustly controlling chaotic PDEs with partial observations
M. Weissenbacher, A. Borovykh and G. Rigas
ICLR Workshop on AI4Differential Equations In Science, 2024

Teaching

I taught week 7 of Dr. Anastasia Borovykh’s course on Mathematical Foundations of Machine Learning at Imperial College London, spring term 2024.

The lecture introduces the idea of contrastive learning, a type of self-supervised learning. It discusses common types of contrastive losses and their interpretations, along with many applications and pitfalls to watch out for.

Take a look at the lecture slides and the accompanying Jupyter notebook, where we implement the SimCLR loss from scratch.