Materials – Machine learning and microscopy

August 3, 2015 — A new technique developed by microscopy and computing experts at Oak Ridge National Laboratory could accelerate advances in materials science and engineering. The team’s approach combines high-performance computing and machine learning algorithms to analyze atomic-scale images and videos from electron and scanning probe microscopes. The near-real time analysis will help researchers extract more chemical and physical information from high-resolution images than previously possible. The ORNL approach focuses on automatically identifying characteristics in “atomic neighborhoods” because local patterns can define a material’s overall properties. This automated data collection lays the framework for building image genomes and libraries to support materials research and design. The team’s study is published in Nature Communications.

---