Introduction

We aim to solve problems in astrophysics and cosmology by combining world-class simulations by Fugaku with AI. In this workshop, the members of three Fugaku projects on astrophysics, earth science, and cosmology will gather together to share research results using simulations and machine learning on a wide range of topics, including large-scale scale structure, galaxy formation, stellar evolution, and planets. We will discuss future AI applications in these fields. In addition to keynote talks and research presentations, we will have a collaborative hack session.

Information

Time: September 11-12 2023
Location: Room 120, University of Tsukuba, Tokyo Campus (access) Slides: link

Schedule:

Day 1 (11 Sep.)

10:30 - 10:35 Opening remarks (Kana Moriwaki)
10:35 - 11:05 Keynote talk (Yusuke Iida, Slides in English, talk in Japanese)
11:05 - 11:15 break
11:15 - 11:45 Keynote talk (Jingjing Shi)
11:45 - 12:00 Overviews on Fugaku projects
(1) Ken Ohsuga
(2) Hideyuki Hotta
(3) Jia Liu

12:00 - 13:30 Lunch break
13:30 - 13:50 Self Introduction (~1 min per person; please put your self-introduction slides on Google Slide)
13:50 - 17:00 Hack (We will divide the participants into 4 groups, evenly distributing people with AI-related experience; the organizer will prepare some tutorials just in case group has no idea where to start; 3-5 min reports from each group in the end)
18:00 ~ Self-organized dinner

Day 2 (12 Sep.)

10:00 - 12:00 Research talks (10 min talk + 5 min Q & A)

  • 10:00 - 10:15 Hiroyuki Masaki “Estimation of thermal convective structure in the solar interior by AI using solar simulations as training data”
  • 10:15 - 10:30 Yuta Asahina “Machine learning estimation of the Eddington tensor for the simulation of accretion disks around black holes”
  • 10:30 - 10:45 Akira Harada “Machine Learning for supernova simulations”
  • 10:45 - 11:00 Keiya Hirashima “Accelerating SN simulations using deep learning toward star-by-star galaxy simulations”
  • 11:00 - 11:15 Kana Moriwaki “Deep learning for line intensity mapping: reconstruction of the large-scale structure of the universe”
  • 11:15 - 11:30 Akira Tokiwa “Spherical Super resolution on HEALPix maps”
  • 11:30 - 11:45 Armijo Joaquin “Weak lensing map modelling using generative methods”
  • 11:45 - 12:00 Hideki Tanimura “DM density reconstruction from galaxy distribution with deep learning”

12:00 - 16:00 Free Lunch + Hack (3-5 min reports from each group in the end)

Notice

  • Eating and drinking are not allowed in the seminar room. Sorry for the inconvenience.
  • You can use eduroam

Language:
Presentations and slides are in English.
The Q&A can be in Japanese if needed.
We will have one introductory talk in Japanese.

Organizers

Kana Moriwaki (UTAP/RESCEU, U Tokyo) - Scientific Chair
Ken Ohsuga (University of Tsukuba) - Local Organizer
Hideyuki Hotta (Nagoya University)
Jia Liu (Kavli IPMU)

Photo

fugakuAI

Acknowledgement

This workshop is supported by Program for Promoting Researches on the Supercomputer Fugaku: Structure and Evolution of the Universe Unraveled by Fusion of Simulation and AI; Grant Number JPMXP1020230406, and JSPS KAKENHI grant No 21H04488.

This page is created by Kateryna Vovk (CD3 research staff).