Design of High T<sub>g</sub> Fluoropolymer using Transfer Learning
발표자
()
초록
내용
We propose a novel machine learning approach to accurately predict fluoropolymers' Tg and demonstrate its potential for designing high Tg copolymers. Firstly, we utilize the QM9 dataset for model pre-training, providing robust molecular representations for subsequent transfer learning on a specialized copolymer dataset. Our pre-trained model expertly encodes complicated molecular structures and general molecular properties using atom-level and global molecular-level features. This feature set is processed via a dual network system that handles a comprehensive molecular descriptor. And we adopt an ensemble approach to deal with model inconsistencies due to limited data. This approach improves the maximum R2 for individual predictions from 0.579 for models trained from scratch to 0.645 for fine-tuned models and 0.696 for the fine-tuned ensemble model. Finally, we can explore a vast chemical space comprising 61 monomers and identify promising candidates for high Tg fluoropolymers.