PairNet: Predicting Correlation Energies with Correlated Features

We proposed a molecular orbital based machine learning model for predicting accurate CCSD(T) correlation energies. The model, named as PairNet, shows excellent transferability on several public data sets using features inspired by pair natural orbitals(PNOs). Though additional computational costs were required to calculate PNOs, the improvement reveals the direct and physical inspired features are essential for a successful deployment of machine learning models in computational chemistry.