Cargando…
Scalable deeper graph neural networks for high-performance materials property prediction
Machine-learning-based materials property prediction models have emerged as a promising approach for new materials discovery, among which the graph neural networks (GNNs) have shown the best performance due to their capability to learn high-level features from crystal structures. However, existing G...
Autores principales: | , , , , , , , |
---|---|
Formato: | Online Artículo Texto |
Lenguaje: | English |
Publicado: |
Elsevier
2022
|
Materias: | |
Acceso en línea: | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9122959/ https://www.ncbi.nlm.nih.gov/pubmed/35607621 http://dx.doi.org/10.1016/j.patter.2022.100491 |
_version_ | 1784711458633809920 |
---|---|
author | Omee, Sadman Sadeed Louis, Steph-Yves Fu, Nihang Wei, Lai Dey, Sourin Dong, Rongzhi Li, Qinyang Hu, Jianjun |
author_facet | Omee, Sadman Sadeed Louis, Steph-Yves Fu, Nihang Wei, Lai Dey, Sourin Dong, Rongzhi Li, Qinyang Hu, Jianjun |
author_sort | Omee, Sadman Sadeed |
collection | PubMed |
description | Machine-learning-based materials property prediction models have emerged as a promising approach for new materials discovery, among which the graph neural networks (GNNs) have shown the best performance due to their capability to learn high-level features from crystal structures. However, existing GNN models suffer from their lack of scalability, high hyperparameter tuning complexity, and constrained performance due to over-smoothing. We propose a scalable global graph attention neural network model DeeperGATGNN with differentiable group normalization (DGN) and skip connections for high-performance materials property prediction. Our systematic benchmark studies show that our model achieves the state-of-the-art prediction results on five out of six datasets, outperforming five existing GNN models by up to 10%. Our model is also the most scalable one in terms of graph convolution layers, which allows us to train very deep networks (e.g., >30 layers) without significant performance degradation. Our implementation is available at https://github.com/usccolumbia/deeperGATGNN. |
format | Online Article Text |
id | pubmed-9122959 |
institution | National Center for Biotechnology Information |
language | English |
publishDate | 2022 |
publisher | Elsevier |
record_format | MEDLINE/PubMed |
spelling | pubmed-91229592022-05-22 Scalable deeper graph neural networks for high-performance materials property prediction Omee, Sadman Sadeed Louis, Steph-Yves Fu, Nihang Wei, Lai Dey, Sourin Dong, Rongzhi Li, Qinyang Hu, Jianjun Patterns (N Y) Article Machine-learning-based materials property prediction models have emerged as a promising approach for new materials discovery, among which the graph neural networks (GNNs) have shown the best performance due to their capability to learn high-level features from crystal structures. However, existing GNN models suffer from their lack of scalability, high hyperparameter tuning complexity, and constrained performance due to over-smoothing. We propose a scalable global graph attention neural network model DeeperGATGNN with differentiable group normalization (DGN) and skip connections for high-performance materials property prediction. Our systematic benchmark studies show that our model achieves the state-of-the-art prediction results on five out of six datasets, outperforming five existing GNN models by up to 10%. Our model is also the most scalable one in terms of graph convolution layers, which allows us to train very deep networks (e.g., >30 layers) without significant performance degradation. Our implementation is available at https://github.com/usccolumbia/deeperGATGNN. Elsevier 2022-04-27 /pmc/articles/PMC9122959/ /pubmed/35607621 http://dx.doi.org/10.1016/j.patter.2022.100491 Text en © 2022 The Author(s) https://creativecommons.org/licenses/by-nc-nd/4.0/This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). |
spellingShingle | Article Omee, Sadman Sadeed Louis, Steph-Yves Fu, Nihang Wei, Lai Dey, Sourin Dong, Rongzhi Li, Qinyang Hu, Jianjun Scalable deeper graph neural networks for high-performance materials property prediction |
title | Scalable deeper graph neural networks for high-performance materials property prediction |
title_full | Scalable deeper graph neural networks for high-performance materials property prediction |
title_fullStr | Scalable deeper graph neural networks for high-performance materials property prediction |
title_full_unstemmed | Scalable deeper graph neural networks for high-performance materials property prediction |
title_short | Scalable deeper graph neural networks for high-performance materials property prediction |
title_sort | scalable deeper graph neural networks for high-performance materials property prediction |
topic | Article |
url | https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9122959/ https://www.ncbi.nlm.nih.gov/pubmed/35607621 http://dx.doi.org/10.1016/j.patter.2022.100491 |
work_keys_str_mv | AT omeesadmansadeed scalabledeepergraphneuralnetworksforhighperformancematerialspropertyprediction AT louisstephyves scalabledeepergraphneuralnetworksforhighperformancematerialspropertyprediction AT funihang scalabledeepergraphneuralnetworksforhighperformancematerialspropertyprediction AT weilai scalabledeepergraphneuralnetworksforhighperformancematerialspropertyprediction AT deysourin scalabledeepergraphneuralnetworksforhighperformancematerialspropertyprediction AT dongrongzhi scalabledeepergraphneuralnetworksforhighperformancematerialspropertyprediction AT liqinyang scalabledeepergraphneuralnetworksforhighperformancematerialspropertyprediction AT hujianjun scalabledeepergraphneuralnetworksforhighperformancematerialspropertyprediction |