# Transfer learning of GW Bethe–Salpeter equation excitation energies

## TL;DR

This paper shows how pretraining machine learning models on DFT and TDDFT data improves predictions of high-fidelity electronic properties with less expensive data.

## Contribution

The novelty is using transfer learning to reduce reliance on costly qsGW and BSE data for accurate electronic-structure predictions.

## Key findings

- Pretraining on DFT and TDDFT improves accuracy for qsGW and BSE predictions.
- Transfer learning reduces the need for expensive high-fidelity data.
- The method works well for larger or chemically distinct molecules.

## Abstract

A persistent challenge in machine learning for electronic-structure calculations is the sharp imbalance between abundant low-fidelity data like (time-dependent) density functional theory [(TD)DFT] results and the scarcity of high-fidelity data like many-body perturbation theory labels. We show that transfer learning provides an effective route to bridge this gap: graph neural networks pretrained on DFT and TDDFT properties can be finetuned with limited qsGW and qsGW-Bethe–Salpeter Equation (BSE) data to yield accurate predictions of quasiparticle and excitation energies. Assessing both full-model and readout-only finetuning across chemically diverse test sets, we find that pretraining improves accuracy, reduces reliance on costly qsGW data, and mitigates large predictive outliers even for molecules larger or chemically distinct from those seen during finetuning. Our results demonstrate that multi-fidelity transfer learning can substantially extend the reach of many-body-level predictions across chemical space.

We show how DFT and TDDFT pretraining improve the accuracy of machine learning-based predictions of qsGW quasiparticle and qsGW-BSE excitation energies. Pretraining furthermore reduces the required amount of expensive qsGW and qsGW-BSE data.