Machine Learning for Defect Analysis in Materials Simulations

  • Construction of Feature Vectors from Materials Structure Data
  • Support Vector Machine for defect classification
  • Greater than 99% accuracy in clasification of point and extended defects
Predicted True Labels
The python-based machine learning module identifies and classifies different point and extended defects in crystal structure data generated by atomistic simulations, by learning information about local atomic structure.
Software Download Links
Source Code
ML Defect Analysis Source Code
Documentation download links

PDF of slides from the MAGICS Workshop on Machine Learning can be downloaded here

Frequently Asked Questions

For frequently asked questions, please click here

Published on December 9th, 2018

Last updated on April 8th, 2021