This document is a practical guide to deep learning in the context of physical simulations, providing hands-on code examples in Jupyter notebooks. It goes beyond standard supervised learning to explore topics such as physical loss constraints, differentiable simulations, and reinforcement learning. The upcoming chapters will cover training networks to infer fluid flow around shapes, using model equations as residuals for training, and interacting with simulators for inverse problems. Throughout the text, different approaches for integrating physical models into deep learning are discussed. The content, maintained by the Physics-based Simulation Group at TUM, emphasizes executable code examples and welcomes feedback for improvement.
https://physicsbaseddeeplearning.org/intro.html