Rapidly predicting Kohn–Sham total energy using data-centric AI

Abstract

Predicting material properties by solving the Kohn-Sham (KS) equation, which is the basis of modern computational approaches to electronic structures, has provided significant improvements in materials sciences. Despite its contributions, both DFT and DFTB calculations are limited by the number of electrons and atoms that translate into increasingly longer run-times. In this work we introduce a novel, data-centric machine learning framework that is used to rapidly and accurately predicate the KS total energy of anatase $${\mathrm{TiO}}_2$$ TiO 2 nanoparticles (NPs) at different temperatures using only a small amount of theoretical data. The proposed framework that we call co-modeling eliminates the need for experimental data and is general enough to be used over any NPs to determine electronic structure and, consequently, more efficiently study physical and chemical properties. We include a web service to demonstrate the effectiveness of our approach.

Structural and Electronic Properties of Two‐Dimensional Materials: A Machine‐Learning‐Guided Prediction

“…Despite its usefulness, DFT computations are bound by the amount of electrons and atoms, which results in steadily greater execution durations. Recently few datacentric ML frameworks show the usefulness of their models for accurately predicting KS energies for quantum mechanical systems [54,55] . Numerous chemical and biological processes take place at or very close to a material's surface, making precise measurements of this area essential in many scientific and engineering fields.…”

“…Recently few datacentric ML frameworks show the usefulness of their models for accurately predicting KS energies for quantum mechanical systems. [54,55] Numerous chemical and biological processes take place at or very close to a material's surface, making precise measurements of this area essential in many scientific and engineering fields. There are few previous studies where it has been shown that ML can accurately predict the surface areas of few systems with a very high accuracy rate.…”

Machine-Learning-Based Detection of Pressure-Induced Faults in Continuous Glucose Monitors

“…Ju et al addressed similar issues in medical imaging by measuring disagreement label noise, and single-target label noise . Future work will include taking a data-centric approach to improving model performance, where higher-resolution labels will be generated to provide richer information from which the models can learn . The soft-labeling technique is commonly adopted to quantify label uncertainty and allow the models to distinguish the sample strongly, weakly, or marginally representing each class. ,− By respecting the intrinsic variation in PISA samples and quantifying the uncertainty in labels using soft-labeling approaches, the model will be able to learn more robust representations of PISA events with higher efficiency.…”

Opportunities and Challenges in Data-Centric AI

“…By breaking down the development process into four building blocks and prioritizing data over conventional process-centric techniques, RoboFlow enables the creation of data-driven AI-enhanced robots. In the field of materials science [105], machine learning is facilitating the prediction of structural features and the discovery of new materials [106]. By utilizing data-centric machine learning methodologies, accurate predictions can be made with minimal theoretical data, improving computational efficiency.…”