| 研究課題/領域番号 |
21K14675
|
|---|
| 研究種目 |
若手研究
|
| 配分区分 | 基金 |
|---|
| 審査区分 |
小区分35010:高分子化学関連
|
|---|
| 研究機関 | 統計数理研究所 |
|---|
研究代表者 |
Wu Stephen 統計数理研究所, 先端データサイエンス研究系, 准教授 (70804186)
|
|---|
| 研究期間 (年度) |
2021-04-01 – 2024-03-31
|
| 研究課題ステータス |
完了 (2023年度)
|
| 配分額 *注記 |
4,550千円 (直接経費: 3,500千円、間接経費: 1,050千円)
2023年度: 1,950千円 (直接経費: 1,500千円、間接経費: 450千円)
2022年度: 1,300千円 (直接経費: 1,000千円、間接経費: 300千円)
2021年度: 1,300千円 (直接経費: 1,000千円、間接経費: 300千円)
|
|---|
| キーワード | polymer informatics / generative models / open source software / ensemble learning / virtual library |
|---|
| 研究開始時の研究の概要 |
I propose to generate collections of polymer candidates with machine learning that will be openly available in a single user-friendly platform, and will serve as a handy starting point for polymer scientists to tackle various design problems of functional polymers along with their expert knowledge.
|
| 研究実績の概要 |
In 2023, we continued the validation of the virtual libraries, as originally planned. We also successfully refined the ensemble of polymer generators within the XenonPy platform, our open source software developed in this project, based on the new measurement results for the synthesized polyimides last fiscal year. Candidates in the virtual library are further filtered using state-of-the-art clustering algorithm and a new model for solubility estimation of polymers in different solvents in order to help the polymer scientists ranking the order for synthesis trials. Our analyses demonstrated that the data-driven polymer design approach based on our generative models is effective on discovery of new functional polymers with desired properties. A total of six new polyimides that were confirmed to exhibit liquid crystalline structure through X-ray diffraction measurement. Our models also continue to contribute to the RadonPy project, which is a collaborative national project between academia and industry to form the largest open calculation database for polymers. This project is just a starting point and we will continue to expand the capacity of our generative models and predictive models for different polymer type and polymer properties to set the groundwork for future innovations in polymer design and materials science.
|
