基于图网络和优化算法的晶体结构预测方法的比较
Comparative study of crystal structure prediction approaches based on a graph network and an optimization algorithm
杨帆 1程观剑 1尹万健2
作者信息
- 1. College of Energy,Soochow Institute for Energy and Materials InnovationS(SIEMIS),and Jiangsu Provincial Key Laboratory for Advanced Carbon Materials and Wearable Energy Technologies,Soochow University,Suzhou 215006,China
- 2. College of Energy,Soochow Institute for Energy and Materials InnovationS(SIEMIS),and Jiangsu Provincial Key Laboratory for Advanced Carbon Materials and Wearable Energy Technologies,Soochow University,Suzhou 215006,China;Shanghai Qi Zhi Institute,Shanghai 200232,China
- 折叠
摘要
本文结合数据库、图神经网络和优化算法,构成了预测晶体结构的一种高效方法.考虑到数据库、图网络架构以及优化算法众多的可选方案,我们需要建立一个基准测试方法,用以评估这些不同方法在性能上的差异.在本研究中,我们从材料数据库中筛选了 100种晶体结构,并建立了一个名为TBCSP的晶体结构预测基准测试,旨在迅速且准确地评估各种晶体结构预测方法的有效性.研究表明,通过结合Materi-als Project数据库、M3GNet架构和贝叶斯优化算法,可以达到高达40％的预测精度.考虑到训练数据的有限性以及元素和晶体结构的多样性,这一发现为通过适度扩充训练数据以进一步提升准确性展示了一条充满希望的路径.
Abstract
The combination of a database,graph neural network,and an optimization algorithm is an effective ap-proach for crystal structure prediction(CSP).Considering that there are multiple options for databases,graph network architectures,and optimization algorithms,a benchmark is required to compare the performances of different ap-proaches.We selected 100 crystal structures from the mate-rials database and established a test benchmark for CSP(TBCSP)aimed at rapidly and accurately assessing the per-formance of various CSP approaches.We found that a com-bination of the Materials Project database,M3GNet architecture,and Bayesian optimization could achieve a pre-diction accuracy of up to 40％.These results are encouraging considering the limited amount of training data,diverse ele-ments and crystal structures.This paper provides a promising way to further enhance the accuracy by properly increasing the training data.
关键词
crystal structure prediction/machine learning/graph network/bayesian optimization/benchmarkKey words
crystal structure prediction/machine learning/graph network/bayesian optimization/benchmark引用本文复制引用
基金项目
国家重点研发计划(2020YFB1506400)
国家自然科学基金(11974257)
Jiangsu Distinguished Young Talent Funding(BK20200003)
DFT calculations were performed at the National Supercomputer Center in Tianjin(TianHe-1A)
出版年
2024
NETL
NSTL
万方数据