Speaker
Yuto Ashida
(The University of Tokyo)
Description
In recent years, the rapid advancement of generative AI, particularly diffusion models, has led to the adoption of new data-driven methods in widely diverse fields—from high-quality image generation to drug discovery and material design. In this talk, I will discuss their theoretical similarities to the theory of the exact renormalization group widely used in statistical and high-energy physics. Specifically, I will introduce an approach based on RG concepts that aim to capture the hierarchical structure of data more efficiently.
Based on K. Masuki and YA, arXiv:2501.09064
| Affiliation | The University of Tokyo |
|---|---|
| Link to paper | https://arxiv.org/abs/2501.09064 |
| Career status | Senior |
Author
Yuto Ashida
(The University of Tokyo)
Co-author
Kanta Masuki
(The University of Tokyo)