2021 Fiscal Year Research-status Report
Towards reaction-network-based reservoir computers as controllers for molecular robots
| Project/Area Number |
20K12061
|
|---|
| Research Institution | Ochanomizu University |
|---|
Principal Investigator |
オベル加藤 ナタナエル お茶の水女子大学, 基幹研究院, 講師 (10749659)
|
|---|
| Co-Investigator(Kenkyū-buntansha) |
GENOT Anthony 東京大学, 生産技術研究所, 国際研究員 (00761975)
|
|---|
| Project Period (FY) |
2020-04-01 – 2023-03-31
|
| Keywords | Molecular Robotics / Reservoir Computing / Thermosensing / Quality-Diversity / PEN DNA Toolbox |
|---|
| Outline of Annual Research Achievements |
During this fiscal year, we adapted a Quality-Diversity algorithm to explore the capabilities of our molecular reservoir computing approach. Such algorithm is designed to put a larger emphasis on the diversity of systems, compared to typical optimization algorithms. Here, we explored the range of various characteristics relevant to reservoir computing (memory capacity, kernel rank, generalization rank) that can be achieved with our approach. We further provide three generation strategies for reservoirs: (a) the optimization of the properties of the random generator, (b) the direct optimization of general purpose reservoir, and (c) a combination of both approaches. We show that each approach can generate reservoirs with different ranges of characteristics, making them thus appropriate for different categories of tasks.
On the experimental side, we implemented a much larger molecular reservoir and showed that it performed as expected. We also generated a large amount of experimental data showing the impact of the temperature (input of our system) on the different molecular elements. Finally, we developed a mathematical model for that system, which fitted particularly well the data. Such model allowed us to get new insights on the dynamics of the reservoir near transitions.
|
| Current Status of Research Progress |
Current Status of Research Progress
1: Research has progressed more than it was originally planned.
Reason
All elements planned for the year (QD implementation, modeling of the impact of temperature, implementation of larger reservoirs) have been done successfully. Moreover, the mathematical model for the molecular reservoir yielded richer dynamics than expected.
In particular, the current results hinted at promising behaviors over a two dimensional reaction diffusion environment. Those results lead us to refine our original plan to focus on such systems next year.
|
| Strategy for Future Research Activity |
Based on the results of this year, we decided to focus more on the implementation of microfluidic systems combined with a temperature range as input. Such system allows us to verify in-vitro the theoretical behavior we obtained. In particular, we expect that the temperature range can act as a differentiation gradient, providing ground for an artificial morphogenesis. Such system can be seen as a specific instance of the original plan, which was supposed to focus on auto-adaptive systems.
We also plan to finalize our mathematical modeling of the impact of temperature on our molecular reservoirs and submit those results as a research article.
|
| Causes of Carryover |
As per the original plan, we will continue implementing molecular reservoir in-vitro. As such, we expect most of the funding to cover the costs of consumables.
Due to the current pandemic, conferences took place online during the 2021 fiscal year, thus reducing traveling costs. Those will be used for additional conference participations and potentially open access publication. In the case traveling is impossible, we will instead use the funding for additional experiments.
|
Research Products
(6 results)
