| Speaker: | Michiko Fujii (UTokyo) |
|---|---|
| Title: | Toward star-by-star Milky-Way galaxy simulations with AI |
| Date (JST): | Wed, Feb 18, 2026, 15:30 - 17:00 |
| Place: | Lecture Hall |
| Abstract: |
Numerical simulations of stellar systems such as globular clusters and galaxies are approaching star-by-star resolution, with individual star particles having realistic stellar masses. With recent supercomputers, it is possible to run a star-by-star simulation of the Milky Way galaxy, but the computational cost is too high to run the model long enough. The main issue is that higher resolution results in smaller timesteps, thereby increasing computational cost. To solve this problem, we are developing a new AI-based scheme. Instead of integrating the time evolution of gas with small timesteps, we use a surrogate model predicted with a machine learning model. With it, we successfully decreased the calculation cost. I will introduce our scheme and preliminary results. |
