Edge Based Architecture for Total Energy Regression Models for Computational Materials Science

  • Kangmo Yeo
  • , Sukmin Jeong
  • , Soo Hyung Kim*
  • *Corresponding author for this work

Research output: Contribution to conferenceConference paperpeer-review

Abstract

It is widely acknowledged that artificial intelligence (AI) technology has been extensively applied and has achieved remarkable advancements in various fields. The field of computational materials science has also embraced AI techniques in diverse ways. Today, computational materials science plays a crucial role in the development of cutting-edge materials, including pharmaceuticals, catalysts, semiconductors, and batteries. One significant task in this field is the regression of the total energy of atomic structures that form various materials. In this study, we propose a modified model architecture aimed at improving the performance of existing total energy regression models. Traditional total energy regression models calculate the total energy by summing the energies of individual nodes represented in the atomic structure graph. However, our approach suggests a modified architecture that not only predicts the energy for nodes but also incorporates energy prediction for edges in the graph. This novel architecture achieved a 3.9% reduction in energy error compared to the base model. Moreover, its simplicity provides the advantage of general applicability to other total energy regression models.

Original languageEnglish
Title of host publicationPattern Recognition - 7th Asian Conference, ACPR 2023, Proceedings
EditorsHuimin Lu, Michael Blumenstein, Sung-Bae Cho, Cheng-Lin Liu, Yasushi Yagi, Tohru Kamiya
PublisherSpringer Science and Business Media Deutschland GmbH
Pages106-112
Number of pages7
ISBN (Print)9783031476648
DOIs
StatePublished - 2023
Event7th Asian Conference on Pattern Recognition, ACPR 2023 - Kitakyushu, Japan
Duration: 2023.11.52023.11.8

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume14408 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference7th Asian Conference on Pattern Recognition, ACPR 2023
Country/TerritoryJapan
CityKitakyushu
Period23.11.523.11.8

Keywords

  • Atomic structure
  • Computational materials science

Quacquarelli Symonds(QS) Subject Topics

  • Computer Science & Information Systems

Fingerprint

Dive into the research topics of 'Edge Based Architecture for Total Energy Regression Models for Computational Materials Science'. Together they form a unique fingerprint.

Cite this