Our new framework can potentially advance the frontiers of expensive and multi-scale simulations, which remain challenging even with the largest supercomputers.
(Left: direct simulation; Right: our framework)

Check out our latest results @NeurIPSConf @ML4PhyS on Dec. 15!

🌌How can we bridge the vast physical scales in computationally expensive physical simulations?

Introducing ASURA-FDPS-ML: The first hybrid framework leveraging surrogate modeling to predict supernova explosions in real-time in galaxy simulations.
📜: ml4physicalsciences.github.io/2024/files/Neu…