Integration of Density Functional Theory with Machine Learning: An Overview

This chapter explores machine learning (ML) integration into density functional theory (DFT) and addresses challenges related to accuracy, computational cost and scalability. It provides a comprehensive overview of different types of ML models and algorithms, followed by discussions on ML-driven improvements to DFT components such as the exchange–correlation functionals, basis sets, self-consistent field and integration grid using various ML techniques. The applications of different ML models, including supervised, unsupervised and semi-supervised learning that replace direct DFT calculations, are also discussed. Case studies highlight the significant impact of ML-DFT integration across various domains, from electronic properties and catalytic reaction predictions to screening materials with the desired properties. Despite these advancements, challenges remain in the ability of some ML models to generalise with new data and scale effectively, highlighting the opportunities for further development and innovation.