I was a Marie Skłodowska-Curie Actions Independent Fellow from June 2020 until May 2022 as part of the European Unions Horizon 2020 research and innovation program under the Marie Skłodowska-Curie grant agreement No. 884741. My project was titled Adapting recurrent neural network algorithms for single molecular break junction analysis.
The main focus of my work was to study the possibility for machine learning to improve experimental and analytical methods in single molecule conductance research. This took two forms: 1) data analysis and 2) simulation. The project depended on a visit to Budapest for a data acquisition portion which proved challenging during the period of the project. To manage this setback, I chose to implement an experimental setup at the CHEM department here at the University of Copenhagen.
Unfortunately, this too faced major setbacks when some of the electronics I ordered were delayed. As a third fallback, I set out to analyze existing data sets. I spent a great deal of time with my colleague, William Bro-Jørgensen, exploring different existing machine learning approaches to obtain an intuition for what approaches might prove fruitful. William authored an interesting review in Chemical Society Reviews which summarized some of our conclusions.
We concluded that single molecule conductance data suffers from a number of challenges that make it extremely difficult for conventional machine learning and neural network methods to be applied effectively and responsibly to it.
Nonetheless, the work I published during my doctorate work in Bern applying principal component analysis to single molecule break junction data suggested to me that simpler, more fundamental mathematical approaches are still useful.
Joseph Hamill 