Title: Searching for Order in Disorder: Structural Indicators in Materials Science Deciphered by Deep Learning
Speaker: Takeshi Kawasaki (Associate Professor, Osaka University D3 Center for Large-Scale Computational Science)
Abstract: In recent years, data science methods such as deep learning have been applied to various fields of natural science, achieving remarkable results in structural prediction, classification, and complementary simulations. However, it has been pointed out that deep learning is often treated as a “black box” where the basis for decisions is opaque, and there are limitations in understanding causality and intrinsic structure, which natural science emphasizes. In this talk, I will introduce a new method for enhancing the interpretability of deep learning models by actively incorporating physics-based knowledge, especially for extracting “feature structures” that govern the anomalous physical properties and dynamic behavior of structures lacking order, such as amorphous solids. Furthermore, we will discuss the applicability of this method not only to physics, but also to the analysis of structure-function relationships in biological and medical fields, such as biological tissues and diseased tissues.
- Date
-
20th June, 2025（Fri,）18:00～20:00
- Venue
-
Held online
- Organizer
-
Co-organizer (HRAM The Japan Society for Industrial and Applied Mathematics, D-DRIVE National Network)
- Participation Fee
-
Free（Advance registration required）
https://www-mmds.sigmath.es.osaka-u.ac.jp/structure/activity/ai_data.php?id=95
- web
-
https://www-mmds.sigmath.es.osaka-u.ac.jp/structure/activity/ai_data.php?id=95
- Contact
-
Takashi Suzuki
suzuki@sigmath.es.osaka-u.ac.jp