Unsupervised machine learning for the analytics of quantum devices
Unsupervised machine learning (ML), and in particular data clustering, is a powerful approach for the analysis of datasets and the identification of characteristic features occurring throughout a dataset. Due to its automated and largely unbiased character, it is gaining popularity across various scientific disciplines as the size of datasets has been steadily increasing over the last years.
Our aim is to develop novel unsupervised machine learning (ML) tools, both in terms of data segmentation, as well as the determination of the number of clusters. Our tools are employed for the classification of scientific datasets such as those acquired using the mechanically controllable break-junction (MCBJ) approach and Raman spectroscopy, both which are routinely used in the field of nanoscience to measure the electrical properties of individual molecules, and to investigate the vibrational properties of materials, respectively. An example of data clustering for a Raman Spectroscopy dataset using one of our ML approaches is shown in Figure 1. Here, a suspended graphene membrane is patterned using focused ion beam. Without beforehand knowledge of the system, the algorithm can identify the areas that received a different He+-exposure.
|Figure 1. Experimental Raman spectroscopy dataset acquired on a Helium beam patterned graphene membrane split into 7 clusters.|
Collaborators in Laboratory: Samuel Singh, Prof. Michel Calame
External partners: Prof. van der Zant (TU Delft), Prof. Klemradt (RWTH Aachen)
Funding: This project is partially funded by The EMPAPOSTDOCS-II programme. The EMPAPOSTDOCS-II programme has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement number 754364. Additional funding is provided by the Swiss National Science Foundation (SNSF) under the Spark project no. 196795 and the FET open project QuIET (no. 767187) and the FET open project QuIET (no. 767187.