A Forecasting System of Computational Time of DFT/TDDFT Calculations under the Multiverse ansatz via Machine Learning and Cheminformatics

TL;DR

This paper introduces a machine learning-based forecasting system that predicts the computational time of DFT/TDDFT calculations using cheminformatics and multiverse interpretation, achieving high accuracy across various molecular configurations.

Contribution

The novel system combines cheminformatics, ML models, and multiverse framework to accurately predict DFT/TDDFT computational times for molecules, including unseen functional and basis set combinations.

Findings

Mean relative errors less than 0.2 for predictions

Effective across different DFT functional and basis set combinations

Utilizes multiverse interpretation to handle diverse calculation scenarios

Abstract

A top-level designed forecasting system for predicting computational times of density-functional theory (DFT)/time-dependent density-functional theory (TDDFT) calculations is presented. The computational time is assumed as the intrinsic property for the molecule. Basing on this assumption, the forecasting system is established using the "reinforced concrete", which combines the cheminformatics, several machine-learning (ML) models, and the framework of many-world interpretation (MWI) in multiverse ansatz. Herein, the cheminformatics is used to recognize the topological structure of molecules, the ML/AI models are used to build the relationships between topology and computational cost, and the MWI framework is used to hold various combinations of DFT functionals and basis sets in DFT/TDDFT calculations. Calculated results of molecules from DrugBank dataset show that 1) it can give quantitative predictions of computational costs, typical mean relative errors can be less than 0.2 for DFT/TDDFT calculations with derivations of 25% using the exactly pre-trained ML models, 2) it can also be employed to various combinations of DFT functional and basis set cases without exactly pre-trained ML models, while only slightly enlarge predicting errors.