摘要
近年来,机器学习作为人工智能的重要支柱,在材料科学尤其是二维材料领域得到广泛应用。图神经网络(Graph Neural Network,GNN)作为一种新兴算法架构,最初在计算机视觉和自然语言处理领域取得成功,现被引入材料研究,通过提取结构特征预测材料性质,能有效处理晶体结构数据。尽管现有技术已能预测二维材料的物化性质,GNN也展示了其在二维材料性能预测方面的强大能力,但仍在预测准确性方面有提升空间。研究使用基于注意力机制的晶体图卷积神经网络模型,针对C2DB(Computational 2D Materials Database)数据库中的二维材料数据进行训练,能够精确预测材料的功函数。
In recent years,machine learning,as a crucial pillar of artificial intelligence,has been widely applied in materials science,particularly in the field of two-dimensional materials.Graph Neural Networks(GNNs),as an emerging algorithmic architecture,initially achieved success in computer vision and natural language processing,and are now being introduced into materials research.By extracting structural features to predict material properties,GNNs can effectively process crystal structure data.Although existing technologies can predict the physicochemical properties of two-dimensional materials,and GNNs have demonstrated strong capabilities in predicting the performance of two-dimensional materials,there is still room for improvement in prediction accuracy.This study employs a crystal graph convolutional neural network model based on an attention mechanism,trained on data from the C2DB database of two-dimensional materials,to accurately predict the work function of materials.
作者
张逸静
孙少瑞
Zhang Yijing;Sun Shaorui(College of Materials Science and Engineering,Beijing University of Technology,Beijing,100124)
出处
《当代化工研究》
2025年第8期33-35,共3页
Modern Chemical Research
关键词
机器学习
图神经网络
图卷积网络
二维材料
材料化工
machine learning
graph neural network
graph convolutional networks
2D materials
materials and chemical engineering
