Prediction of Runtime Parameters of Parallel Chemistry Applications via Active and Generative Learning
Researchers have developed two machine learning approaches to predict runtime parameters for parallel chemistry applications. These methods combine active and generative learning with gradient boosted regression trees, achieving a mean absolute percentage error as low as 0.023 and a coefficient of determination of 99.9% on Coupled-Cluster with Singles and Doubles computations. When active learning is used to address limited training data, the models achieve a MAPE of approximately 0.2 with only 20-25% of the original dataset. AI