Title: All the Cool Stuff - Generative Machine Learning for Partial Differential Equations
Speaker: Andrew Millard
Affiliation: Linköping University
Room: Å90101
Time: 14:00-15:00
Abstract:
We have all probably seen the stuff generative models have been providing. From large language models to generating images, videos, audio, and that's just the tip of the iceberg. It turns out the same ideas that let models draw pictures can also help us solve serious scientific problems. In this talk, we’ll take a friendly tour of generative machine learning (with extra love for diffusion models) and see how these tools can crack inverse problems—the kind that show up when we know what happens, but not why. Then we’ll jump into the world of partial differential equations (PDEs): the math that explains heat, fluids, waves, materials, and just about every physical system. Using generative models, we can solve PDEs faster, infer hidden quantities, and explore entirely new designs and materials. The punchline? Once you can solve PDEs efficiently, you unlock so much other cool stuff across physics, chemistry, and biology.