Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals
Graph Networks as a Universal Machine Learning Framework for Molecules and Crystals
复制标题
DOI：
10.1021/acs.chemmater.9b01294
复制
发表时间：
2019-05-14
影响因子：
8.6
通讯作者：
Ong, Shyue Ping
中科院分区：
文献类型：
--
作者：
Chen, Chi;Ye, Weike;Ong, Shyue Ping
图网络是一种新的机器学习（ML）范式，支持关系推理和组合泛化。在这里，我们开发了通用的Materials Graph Network（MEGNet）模型，用于分子和晶体的准确属性预测。我们证明了MEGNet模型在QM9分子数据集的13个属性中的11个中优于先前的ML模型，例如SchNet。同样，我们表明，在材料项目中，在类似于60000个晶体上训练的MEGNet模型在预测晶体的形成能、带隙和弹性模量方面大大优于先前的ML模型，在更大的数据集上实现了比密度泛函理论更好的准确性。我们提出了两种新的策略来解决材料科学和化学中常见的数据限制。首先，我们展示了一种物理直观的方法，通过将温度，压力和熵作为全局状态输入，将0 K和室温下的内能，焓和吉布斯自由能的四个独立的分子MEGNet模型统一为单个自由能MEGNet模型。其次，我们证明了MEGNet模型中的学习元素嵌入编码了周期性的化学趋势，并且可以从在较大数据集（形成能）上训练的属性模型中转移学习，以改进具有少量数据（带隙和弹性模量）的属性模型。
Graph networks are a new machine learning (ML) paradigm that supports both relational reasoning and combinatorial generalization. Here, we develop universal MatErials Graph Network (MEGNet) models for accurate property prediction in both molecules and crystals. We demonstrate that the MEGNet models outperform prior ML models such as the SchNet in 11 out of 13 properties of the QM9 molecule data set. Similarly, we show that MEGNet models trained on similar to 60 000 crystals in the Materials Project substantially outperform prior ML models in the prediction of the formation energies, band gaps, and elastic moduli of crystals, achieving better than density functional theory accuracy over a much larger data set. We present two new strategies to address data limitations common in materials science and chemistry. First, we demonstrate a physically intuitive approach to unify four separate molecular MEGNet models for the internal energy at 0 K and room temperature, enthalpy, and Gibbs free energy into a single free energy MEGNet model by incorporating the temperature, pressure, and entropy as global state inputs. Second, we show that the learned element embeddings in MEGNet models encode periodic chemical trends and can be transfer-learned from a property model trained on a larger data set (formation energies) to improve property models with smaller amounts of data (band gaps and elastic moduli).