[18a-A21-1~11] 23.1 Joint Session N "Informatics"

11:45 AM - 12:00 PM JST(2:45 AM - 3:00 AM UTC)

[18a-A21-11] Support for all stable elements and improved robustness of universal neural network potential PFP

〇So Takamoto¹, Chikashi Shinagawa¹ (1.PFN)

Keywords:

Machine Learning,Interatomic Potential,AI for Science

The universal interatomic potential PFP being developed by the authors is characterized by its versatility in handling arbitrary combination of elements in a single model. The latest PFP supports 96 elements, covering all elements that stably exist on earth. The dataset size has reached approximately 60 million structures. In addition, the robustness of the extrapolated regions has been improved.