Presentation given at the 86th The Japan Society of Applied Physics autumn meeting 2025

We participated inthe 86th The Japan Society of Applied Physics autumn meeting 2025 ,held at Meijo University's Amashiro Campus from September 7th to September 11th, 2025.

 

Presentation from Matlantis

Three members of Matlantis gave poster presentations and oral presentations.

 

Title: Interpretation of atomic-level predictions combining PFP descriptors and Shapley values [8p-P11-6]
Presenter: Bon Cho

Abstract: In recent years, materials informatics, which uses machine learning to develop materials, has been attracting attention. However, many machine learning models are black boxes, and the basis for their predictions is unclear, hindering the acquisition of new scientific knowledge. To address this issue, this study proposes an interpretation method using atomic feature PFP descriptors and Shapley values that can be calculated from the general-purpose machine learning atomic force potential PFP.

Copyright (2025) The Japan Society of Applied Physics

 

The contents of this presentation have been made available as an example for Matlantis users to use.

 

Title: Search for Stable Structures of Magnesium-Based Titanium Nitride Using the General-Purpose Machine Learning Potential PFP[9p-S301-7]

Presenter:Yuta Aoki

Summary:Magnéli-based titanium oxynitrides are compounds obtained by substituting a certain percentage of oxygen atoms in the Magnéli phase of titanium oxides, with the total number of possible substitutions reaching tens of thousands. It is unrealistic to use first-principles calculations to screen for stable structures from this vast number of candidate structures, so in this study we performed this calculation using the general-purpose machine learning potential PFP. We also analyzed what structural features stabilize the structure.

Copyright (2025) The Japan Society of Applied Physics

 

There were many questions and answers, making for a lively session.

 

Matlantis was also introduced as a sponsored talk in the following session:

 

Analysis and application of Atomic Layer Process (ALP) (2) [7a-S103-1~7]
Presenter: Yusuke Asano

It was a valuable opportunity to speak directly with many people. Thank you very much!

 

Latest Articles

NEW

Learning AI materials simulation to accelerate research at the Fukui Kenichi Memorial Research Center, Kyoto University - Working with ENEOS to design the "best CO2 adsorbent"

Learning AI materials simulation to accelerate research at the Fukui Kenichi Memorial Research Center, Kyoto University - Working with ENEOS to design the "best CO2 adsorbent"

NEW

Explainer : Why Did the AI Predict That ? Uncovering Atomic-Level Interpretability through PFP Descriptors and Shapley Values

Bon Cho Bon Cho

Materials Informatics Explainer computational chemistry

Explainer : Why Did the AI Predict That ? Uncovering Atomic-Level Interpretability through PFP Descriptors and Shapley Values

Writing SMILES from scratch

Bon Cho Bon Cho

Explainer computational chemistry

Writing SMILES from scratch

Nagoya University × Matlantis Case Study:“Advanced Experiments for Frontier Technologies and Sciences” —A Four-Day Intensive Course That Sparked Experimental Students’ Curiosity Through AI Simulation

Interview computational chemistry

Nagoya University × Matlantis Case Study:“Advanced Experiments for Frontier Technologies and Sciences” —A Four-Day Intensive Course That Sparked Experimental Students’ Curiosity Through AI Simulation

Introduction to Machine Learning Interatomic Potentials (MLIPs): A Game Changer in Materials Simulation

Yoshitaka Yamauchi Masataka Yamauchi

Machine Learning Interatomic Potentials Explainer

Introduction to Machine Learning Interatomic Potentials (MLIPs): A Game Changer in Materials Simulation