Published September 2, 2025
| Version v1
Software
Open
Code and Data for Generative AI-Powered Inverse Design for Tailored Narrowband Molecular Emitters
Authors/Creators
-
1.
Nanjing University
-
2.
Tsinghua University
Description
Introduction
This project proposes Markov Emission MOlecular Sampling (MEMOS) for the inverse design of OLED molecules. We achieve high inverse design success rates for OLED molecules with target wavelengths and narrow FWHM.
Data availability
- Initial molecule data:
oled_data_new/initial_data.csv - Training data for each iteration:
oled_data_new/unimol_train_data
Run
Run the command outside of the project directory:
bash train_mcmc_oled.shA GPU is required to run this script. All our experiments were conducted using an NVIDIA 3090ti GPU.
Acknowledgements
Our implementation builds upon the MARS codebase, and we acknowledge their excellent contribution.