Congratulations to Lars Blumenthal for winning the third place in kaggle’s data science competition “Nomad2018 Predicting Transparent Conductors”.
The goal of this competition was to develop machine learning models for the prediction of two materials properties, namely the formation energy, which is an indication of the stability of a new material, and the electronic band gap, which determines a material’s transparency over the visible range. The developed models can potentially facilitate the discovery of new transparent conductors and allow for advancements in (opto)electronic technologies.
Inspired by Jacek Golebiowski, who made valuable contributions to the final solution, Lars used a smooth overlap of atomic positions (SOAP) based descriptor developed by Barto?k et al. [1, 2] to encode information about the crystal structure of the transparent conductive oxides that were studied in this competition. These SOAP features were then used to teach a Neural Network to predict the desired materials properties.
Details about the competition and the top three submissions can be found here.
[1] A. P. Bartók, R. Kondor, and G. Csányi, Physical Review B, vol. 87, no. 18, May 2013.
[2] S. De, A. P. Bartók, G. Csányi, and M. Ceriotti, Physical Chemistry Chemical Physics, vol. 18, no. 20, pp. 13754–13769, 2016.
Article text (excluding photos or graphics) © Imperial College London.
Photos and graphics subject to third party copyright used with permission or © Imperial College London.
Reporter
Miranda Toora
Department of Materials
