Molekuláris elektronikai mérések elemzése modern gépi tanulási és adatbányászati módszerekkel

Nyomtatóbarát változatNyomtatóbarát változat
Cím angolul: 
Analysis of molecular electronic experiments with modern machine learning and data minig techniques
MSc diplomamunka téma - nanotechnológia és anyagtudomány
Dr. Halbritter András Ernő
Email cím:
BME Fizika Tanszék
egyetemi tanár, tanszékvezető
Balogh Nóra
Fizikus MSc - nanotechnológia és anyagtudomány

Computer programing and data analysis skills, basic knowledge of solid state physics and materials science.


The field of molecular electronics, i.e. the application of single molecules as the building blocks of electronic equipment is among the research directions searching for novel information technologies. There are already well-established techniques to form single molecule nanowires or even more complex single molecule devices.


The so-called break junction technique is a widely used method for the statistical analysis of single molecule nanowires. The typical datasets include ten thousands of conductance vs. electrode separation traces with hundreds of thousands of datapoints on each trace. Our group has developed some statistical methods (like 2D correlation analysis) that can be effectively used to analyze these data, and to understand the temporal evolution of the few atom and single molecule nanowires in more depth than the information supplied by simple conductance histograms. Within the diploma thesis the applicant would broaden our statistical analysis toolbox towards the application of modern machine learning and data mining techniques. This would include the analysis of single molecule break junction data with neural network algorithms, principal component analysis methods and advanced clustering methods. These techniques could help to identify patterns in the data, which are hardly recognized by traditional approaches, and thus would provide a better understanding of the underlying physical processes. Concerning the neural network based analysis, it is an important goal to apply simple enough neural networks, such that one can get insight to the decision taking mechanism of the network.


Hozzáférés nincs korlátozva