Summary
Molecular Electronics Artificial Neural Networks (MEANN) will adapt for the first time a recurrent neural network (RNN) to address complex multivariate correlation questions that arise in single molecular break junction (SMBJ) experiments. The hypothesis is that a RNN will be better than a human at identifying relationships between nanoscopic geometry changes of the junctions and the measured variables in SMBJ data sets, with little or no human bias. These improvements in the data analysis approach will allow researchers to address many of the present problems in SMBJ research, most notably reproducibility and bridging the theory-experiment gap. The proposal has three objectives to implement this goal. I will: (1) generate simulated SMBJ data and use this simulated data to train a RNN to sort SMBJ data into classes with unique and significant features in the data; (2) measure large sets of experimental data while on secondment and apply the trained RNN to the experimental data to sort the experimental data into the classes the RNN has already identified in the simulated data; and (3) derive a deeper understanding of the relationships between the physical processes involved in the break junction, and the observable variables of the experiment. MEANN maximizes my development as a researcher by exposing me to three important opportunities: (1) a world class theoretical chemistry group where I will learn computational and management skills necessary for my future as a researcher, (2) new experimental physics techniques while on secondment, and (3) planning an Applied RNN Summit where I will network with industry leaders in RNN development, share my expertise with peers, and prepare teaching materials to introduce my research to students. As a result of MEANN, researchers will have new tools to generate simulated SMBJ data, analyse their experimental data quickly and objectively, and answer important questions in condensed matter physics and physical chemistry.
Unfold all
/
Fold all
More information & hyperlinks
| Web resources: | https://cordis.europa.eu/project/id/884741 |
| Start date: | 01-06-2020 |
| End date: | 31-05-2022 |
| Total budget - Public funding: | 207 312,00 Euro - 207 312,00 Euro |
Cordis data
Original description
Molecular Electronics Artificial Neural Networks (MEANN) will adapt for the first time a recurrent neural network (RNN) to address complex multivariate correlation questions that arise in single molecular break junction (SMBJ) experiments. The hypothesis is that a RNN will be better than a human at identifying relationships between nanoscopic geometry changes of the junctions and the measured variables in SMBJ data sets, with little or no human bias. These improvements in the data analysis approach will allow researchers to address many of the present problems in SMBJ research, most notably reproducibility and bridging the theory-experiment gap. The proposal has three objectives to implement this goal. I will: (1) generate simulated SMBJ data and use this simulated data to train a RNN to sort SMBJ data into classes with unique and significant features in the data; (2) measure large sets of experimental data while on secondment and apply the trained RNN to the experimental data to sort the experimental data into the classes the RNN has already identified in the simulated data; and (3) derive a deeper understanding of the relationships between the physical processes involved in the break junction, and the observable variables of the experiment. MEANN maximizes my development as a researcher by exposing me to three important opportunities: (1) a world class theoretical chemistry group where I will learn computational and management skills necessary for my future as a researcher, (2) new experimental physics techniques while on secondment, and (3) planning an Applied RNN Summit where I will network with industry leaders in RNN development, share my expertise with peers, and prepare teaching materials to introduce my research to students. As a result of MEANN, researchers will have new tools to generate simulated SMBJ data, analyse their experimental data quickly and objectively, and answer important questions in condensed matter physics and physical chemistry.Status
CLOSEDCall topic
MSCA-IF-2019Update Date
28-04-2024
Images
No images available.
Geographical location(s)
