This repository contains the code and data used for training generative models in the conditional and unconditional design of polymer structures. PolyGen provides tools for:
Training generative machine learning models,
Example LAMMPS input files to assess polymer structures using molecular dynamics (MD),
Evaluating the generated polymer electrolytes in terms if 6 defined metrics.
The framework has been developed as part of research into accelerating polymer electrolyte discovery and used in the following research papers:
-
Yang, Zhenze, et al. "De novo design of polymer electrolytes with high conductivity using gpt-based and diffusion-based generative models." arXiv preprint arXiv:2312.06470 (2023).
-
Khajeh, Arash, et al. "A self-improvable polymer discovery framework based on conditional generative model." arXiv preprint arXiv:2312.04013 (2023).