Generative Models from Differential Equations: Building a Fast Digital Twin
In this hands-on hackathon, you'll learn to build a generative surrogate model that mimics a heavy physics simulation 1000x faster, using Flow Matching, a modern and elegant algorithm based on learning noise-to-data mappings through ODEs. The focus is didactic rather than competitive: you'll work through scaffolded notebooks in Python and PyTorch, implementing the core physics logic and loss functions step by step while we handle the boilerplate. Prior experience with Python and PyTorch is recommended. No GPU required, everything runs on your laptop or Colab. You are welcome to attend as a small group if you prefer.
For more info, please contact francesco.vaselli@pi.infn.it