Hands-on Deep Learning for Materials Science: Convolutional Networks and Variational Autoencoders

By Vinay Hegde1; Alejandro Strachan2

1. Citrine Informatics, Sunnyvale, CA 2. Materials Engineering, Purdue University, West Lafayette, IN

Published on


Run the Tool: Hands-on Deep Learning for Materials This tutorial introduces deep learning techniques such as convolutional neural networks and variational auto encoders from a materials standpoint. Using a solubility dataset, you will learn how to pre-process data and train a convolutional neural network to predict the solubility of polymers. You will also explore variational autoencoders in the context of the solubility dataset.

The tool used in this demonstration is Hands-on Deep Learning for Materials.


Vinay is a Postdoctoral Fellow at Citrine Informatics, where he works with external collaborators in universities, national labs, and other companies on interesting materials informatics problems. Vinay received his PhD in Materials Science and Engineering from Northwestern University, where he co-developed and maintained the Open Quantum Materials Database (or OQMD), one of the largest open databases of calculated materials properties. His thesis research focused on developing data-driven methods for new materials discovery and knowledge extraction.

Cite this work

Researchers should cite this work as follows:

  • Vinay Hegde, Alejandro Strachan (2020), "Hands-on Deep Learning for Materials Science: Convolutional Networks and Variational Autoencoders," https://nanohub.org/resources/34474.

    BibTex | EndNote