Magpie Manual

What is Magpie?

Magpie is an extensible platform for using machine learning to predict the properties of materials.

Magpie is also an acronym for “Material-Agnostic Platform for Informatics and Exploration”, and is named after an intelligent bird.

Reasons to use Magpie

1. Ability to create, test, and use machine learning models all in the same program
2. Interactive text-based scripting interface
3. Easy sharing of models in a machine-independent format

Getting Magpie

Magpie is open source, and is both hosted on BitBucket and available in compiled form on this site. The distribution of Magpie includes:

• All libraries necessary to run Magpie
• This manual, and documentation for the Java interface
• Example scripts for several common tasks

Magpie is written entirely in Java, which allows it to be run on most computers without modification. As long as the Java 8 Runtime Environment is installed on your computer, Magpie should run without issue.

Some features of Magpie require Python and Python libraries to run. See Installation for more details.

Learning About Magpie

If you are brand new to Magpie, you might want to read the tutorial section of this manual and then review a few example scripts:

• Tutorial

• Compiling Magpie

• Text Interface

• Example Scripts

• Variable Types

• Introduction to Java Interface

• Java API Documentation

• How to Contribute

Additionally, you may also want to read the following article, which provides a short introduction to using machine learning to predict new materials and explains the science behind one of methods available through Magpie.

Ward, L., et. al.. npj Computational Materials. 2 (2016), 16028.