No abstract

Obtaining accurate estimates of machine learning model uncertainties on newly predicted data is essential for understanding the accuracy of the model and whether its predictions can be trusted. A common approach to such uncertainty quantification is to estimate the variance from an ensemble of models, which are often generated by the generally applicable bootstrap method. In this work, we demonstrate that the direct bootstrap ensemble standard deviation is not an accurate estimate of uncertainty but that it can be simply calibrated to dramatically improve its accuracy. We demonstrate the effectiveness of this calibration method for both synthetic data and numerous physical datasets from the field of Materials Science and Engineering. The approach is motivated by applications in physical and biological science but is quite general and should be applicable for uncertainty quantification in a wide range of machine learning regression models.

Irradiation increases the yield stress and embrittles light water reactor (LWR) pressure vessel steels. In this study, we demonstrate some of the potential benefits and risks of using machine learning models to predict irradiation hardening extrapolated to low flux, high fluence, extended life conditions. The machine learning training data included the Irradiation Variable for lower flux irradiations up to an intermediate fluence, plus the Belgian Reactor 2 and Advanced Test Reactor 1 for very high flux irradiations, up to very high fluence. Notably, the machine learning model predictions for the high fluence, intermediate flux Advanced Test Reactor 2 irradiations are superior to extrapolations of existing hardening models. The successful extrapolations showed that machine learning models are capable of capturing key intermediate flux effects at high fluence. Similar approaches, applied to expanded databases, could be used to predict hardening in LWRs under life-extension conditions.

Flash points of organic molecules play an important role in preventing flammability hazards and large databases of measured values exist, although millions of compounds remain unmeasured. To rapidly extend existing data to new compounds many researchers have used quantitative structure-property relationship (QSPR) analysis to effectively predict flash points. In recent years graph-based deep learning (GBDL) has emerged as a powerful alternative method to traditional QSPR. In this paper, GBDL models were implemented in predicting flash point for the first time. We assessed the performance of two GBDL models, messagepassing neural network (MPNN) and graph convolutional neural network (GCNN), by comparing against 12 previous QSPR studies using more traditional methods. Our result shows that MPNN both outperforms GCNN and yields slightly worse but comparable performance with previous QSPR studies. The average R ! and Mean Absolute Error (MAE) scores of MPNN are, respectively, 2.3% lower and 2.0K higher than previous comparable studies. To further explore GBDL models, we collected the largest flash point dataset to date, which contains 10575 unique molecules. The optimized MPNN gives a test data R ! of 0.803 and MAE of 17.8 K on the complete dataset. We also extracted 5 datasets from our integrated dataset based on molecular types (acids, organometallics, organogermaniums, organosilicons, and organotins) and explore the quality of the model in these classes.