Gavin Nop, Micah Mundy, Durga Paudyal, Jonathan Smith
{"title":"利用新型忠实机器学习嵌入预测量子材料特性","authors":"Gavin Nop, Micah Mundy, Durga Paudyal, Jonathan Smith","doi":"arxiv-2407.18388","DOIUrl":null,"url":null,"abstract":"Machine Learning (ML) is accelerating the progress of materials prediction\nand classification, with particular success in CGNN designs. While classical ML\nmethods remain accessible, advanced deep networks are still challenging to\nbuild and train. We introduce two new adaptations and refine two existing ML\nnetworks for generic crystalline quantum materials properties prediction and\noptimization. These new models achieve state-of-the-art performance in\npredicting TQC classification and strong performance in predicting band gaps,\nmagnetic classifications, formation energies, and symmetry group. All networks\neasily generalize to all quantum crystalline materials property predictions. To\nsupport this, full implementations and automated methods for data handling and\nmaterials predictions are provided, facilitating the use of deep ML methods in\nquantum materials science. Finally, dataset error rates are analyzed using an\nensemble model to identify and highlight highly atypical materials for further\ninvestigations.","PeriodicalId":501211,"journal":{"name":"arXiv - PHYS - Other Condensed Matter","volume":"44 1","pages":""},"PeriodicalIF":0.0000,"publicationDate":"2024-07-25","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Predicting quantum materials properties using novel faithful machine learning embeddings\",\"authors\":\"Gavin Nop, Micah Mundy, Durga Paudyal, Jonathan Smith\",\"doi\":\"arxiv-2407.18388\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"Machine Learning (ML) is accelerating the progress of materials prediction\\nand classification, with particular success in CGNN designs. While classical ML\\nmethods remain accessible, advanced deep networks are still challenging to\\nbuild and train. We introduce two new adaptations and refine two existing ML\\nnetworks for generic crystalline quantum materials properties prediction and\\noptimization. These new models achieve state-of-the-art performance in\\npredicting TQC classification and strong performance in predicting band gaps,\\nmagnetic classifications, formation energies, and symmetry group. All networks\\neasily generalize to all quantum crystalline materials property predictions. To\\nsupport this, full implementations and automated methods for data handling and\\nmaterials predictions are provided, facilitating the use of deep ML methods in\\nquantum materials science. Finally, dataset error rates are analyzed using an\\nensemble model to identify and highlight highly atypical materials for further\\ninvestigations.\",\"PeriodicalId\":501211,\"journal\":{\"name\":\"arXiv - PHYS - Other Condensed Matter\",\"volume\":\"44 1\",\"pages\":\"\"},\"PeriodicalIF\":0.0000,\"publicationDate\":\"2024-07-25\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"arXiv - PHYS - Other Condensed Matter\",\"FirstCategoryId\":\"1085\",\"ListUrlMain\":\"https://doi.org/arxiv-2407.18388\",\"RegionNum\":0,\"RegionCategory\":null,\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"\",\"JCRName\":\"\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"arXiv - PHYS - Other Condensed Matter","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/arxiv-2407.18388","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
摘要
机器学习(ML)正在加速材料预测和分类的发展,其中 CGNN 设计尤为成功。虽然经典的 ML 方法仍然可以使用,但高级深度网络的构建和训练仍然具有挑战性。我们为通用晶体量子材料性能预测和优化引入了两个新的适应性模型,并改进了两个现有的 ML 网络。这些新模型在预测 TQC 分类方面达到了最先进的性能,在预测带隙、磁性分类、形成能和对称组方面表现出色。所有网络都能轻松地通用于所有量子晶体材料的属性预测。为了支持这一点,我们提供了数据处理和材料预测的完整实现和自动化方法,从而促进了深度 ML 方法在量子材料科学中的应用。最后,使用集合模型分析了数据集错误率,以识别并突出高度非典型材料,供进一步研究。
Predicting quantum materials properties using novel faithful machine learning embeddings
Machine Learning (ML) is accelerating the progress of materials prediction
and classification, with particular success in CGNN designs. While classical ML
methods remain accessible, advanced deep networks are still challenging to
build and train. We introduce two new adaptations and refine two existing ML
networks for generic crystalline quantum materials properties prediction and
optimization. These new models achieve state-of-the-art performance in
predicting TQC classification and strong performance in predicting band gaps,
magnetic classifications, formation energies, and symmetry group. All networks
easily generalize to all quantum crystalline materials property predictions. To
support this, full implementations and automated methods for data handling and
materials predictions are provided, facilitating the use of deep ML methods in
quantum materials science. Finally, dataset error rates are analyzed using an
ensemble model to identify and highlight highly atypical materials for further
investigations.