Databases and graph libraries for machine learning of material science

• These are databases and demo codes for the following paper
• "Integrating multiple materials science projects in a single neural network"
• Published in "Communications Materials"

Install

• conda env create -f=requirements.yaml
• Checked with Ubuntu 18.04.5 LTS
  • Making the environment seems to fail on Windows and Mac (ResolvePackageNotFound error).

Usage

• You can view databases used in the study in a "database" folder
• You can run jupyter notebook codes for the demo codes in a "codes" folder (prediction of conductivity of PEDOT-PSS from texts, multitask learning of multiple databases, inverse problems, and control experiments against the graph approach)

Author

• Kan Hatakeyama-Sato and Kenichi Oyaizu
• Waseda University
• https://kanhatakeyama.github.io/
• http://www.appchem.waseda.ac.jp/~polymer/PolymerDatabase/

Thank you!