Uncertainty estimation in deep learning‐based property models: Graph neural networks applied to the critical properties

Deep learning and graph-based models have gained popularity in various life scienceapplications such as property modeling, achieving state-of-the-art performance.However, the quantification of prediction uncertainty in these models is less studiedand is not applied in the low dataset size regime, which characterizes many chemicalengineering-related molecular properties. In this work, we apply two graph-basedmodels to model the critical- temperature, pressure, and volume and apply threetechniques (the bootstrap, the ensemble, and the dropout) to quantify the predictionuncertainty. The overall model confidence is evaluated using the coverage. Theresults suggest that graph-based models perform better compared with currentgroup-contribution-based property modeling techniques while eliminating thetedious task of developing molecular descriptors.

Recommended citation: Aouichaoui, A. R., Mansouri, S. S., Abildskov, J., & Sin, G. (2022). Uncertainty estimation in deep learning‐based property models: Graph neural networks applied to the critical properties. AIChE Journal, 68(6), e17696. https://doi.org/10.1002/aic.17696