| 研究課題/領域番号 |
20K12061
|
|---|
| 研究機関 | お茶の水女子大学 |
|---|
研究代表者 |
オベル加藤 ナタナエル お茶の水女子大学, 基幹研究院, 助教 (10749659)
|
|---|
| 研究分担者 |
GENOT Anthony 東京大学, 生産技術研究所, 国際研究員 (00761975)
|
|---|
| 研究期間 (年度) |
2020-04-01 – 2023-03-31
|
| キーワード | Molecular Robotics / Reservoir Computing / PEN DNA toolbox |
|---|
| 研究実績の概要 |
Following the plan for the year, we designed and implemented a temperature-based input mechanism for molecular reservoir computing. Using temperature allowed us to interact with the system while keeping it chemically closed, a crucial step to use the reservoir computing approach with standard laboratory equipment.
We implemented the reservoir with a robust molecular oscillator, subjecting it to sudden temperature variations and monitoring its response with fluorescent reporters. We then trained in-silico neural networks on the fluorescence traces to predict the inputted temperature profiles. We reached an average of 87% accuracy for a single layer and 91% for two layers, showing the potential of such reservoir.
|
| 現在までの達成度 (区分) |
現在までの達成度 (区分)
1: 当初の計画以上に進展している
理由
All elements planned for the year (design of the reservoir system, implementation, and experimental validation) have been done successfully. Furthermore, the accuracy of the reservoir on the delay task was better than originally expected, reaching up to 91% accuracy. The results have been published as part of the proceedings of the ALIFE conference.
As such, we were able to move on faster than expected to the next step of the project, involving more complex molecular systems.
|
| 今後の研究の推進方策 |
We will follow the original plan for the second year: we will combine the current thermal model with our Quality-Diversity algorithm and use it to propose better reservoirs with respects to benchmark tasks such as the delay. Additionally to the original plan, we expect to also measure direct characteristics from the reservoirs, such as kernel rank or memory capacity, to get a better understanding of their potential.
Promising systems will be implemented and experimentally validated. Results generated will be used to fit parameters of the model using an optimization technique we previously proposed.
Finally, we plan to implement complex input patterns in-vitro to further test the capabilities of the systems.
|
| 次年度使用額が生じた理由 |
Most of the cost incurred during the next fiscal year will cover the price of consumables (DNA, enzymes, buffer) used in experiments. The remaining budget will cover the cost of conference registration and possible publication fees of our results.
|
