BestPractices

Things that you should (and should not) do in your Materials Informatics research.

This is a repository containing the relevant Python code and Jupyter notebooks to the publication "Machine Learning for Materials Scientists: An Introductory Guide toward Best Practices".

These notebooks are included to illustrate a hypothetical Machine Learning project in materials science created following best practices. The goal of this project is to predict the heat capacity of materials given a chemical composition and condition (the measurement temperature).

To read the main publication for which these notebooks are made, please see:

Wang, Anthony Yu-Tung; Murdock, Ryan J.; Kauwe, Steven K.; Oliynyk, Anton O.; Gurlo, Aleksander; Brgoch, Jakoah; Persson, Kristin A.; Sparks, Taylor D., Machine Learning for Materials Scientists: An Introductory Guide toward Best Practices, Chemistry of Materials 2020, 32 (12): 4954–4965. DOI: 10.1021/acs.chemmater.0c01907.

Table of Contents

• How to cite
• Installation
• Opening the Jupyter notebooks
• Using the Jupyter notebooks
• Julia implementation via Pluto.jl

How to cite

Please cite the following work if you choose to adopt or adapt the methods mentioned in this Methods/Protocols article:

Wang, Anthony Yu-Tung; Murdock, Ryan J.; Kauwe, Steven K.; Oliynyk, Anton O.; Gurlo, Aleksander; Brgoch, Jakoah; Persson, Kristin A.; Sparks, Taylor D., Machine Learning for Materials Scientists: An Introductory Guide toward Best Practices, Chemistry of Materials 2020, 32 (12): 4954–4965. DOI: 10.1021/acs.chemmater.0c01907.

Citation in BibTeX format:

@article{Wang2020bestpractices,
    author = {Wang, Anthony Yu-Tung and Murdock, Ryan J. and Kauwe, Steven K. and Oliynyk, Anton O. and Gurlo, Aleksander and Brgoch, Jakoah and Persson, Kristin A. and Sparks, Taylor D.},
    year = {2020},
    title = {Machine Learning for Materials Scientists: An Introductory Guide toward Best Practices},
    url = {https://doi.org/10.1021/acs.chemmater.0c01907},
    pages = {4954--4965},
    volume = {32},
    number = {12},
    issn = {0897-4756},
    journal = {Chemistry of Materials},
    doi = {10.1021/acs.chemmater.0c01907}
}

Installation

This repositories hosts a series of Jupyter notebooks, which run on the Python programming language. Please follow the below steps to get started with using these notebooks.

Clone or download this GitHub repository

Do one of the following:

• Clone this repository to a directory of your choice on your computer.
• Download an archive of this repository and extract it to a directory of your choice on your computer.

Install dependencies via Anaconda:

1. Download and install Anaconda.
2. Navigate to the project directory (from above).
3. Open Anaconda prompt in this directory.
4. Run the following command from Anaconda prompt to automatically create an environment from the conda-env.yml file:
   • conda env create --file conda-env.yml
5. Run the following command from Anaconda prompt to activate the environment:
   • conda activate bestpractices (bestpractices is the name of the environment)

For more information about creating, managing, and working with Conda environments, please consult the relevant help page.

Install dependencies via pip:

Open conda-env.yml and pip install all of the packages listed there. We recommend that you create a separate Python environment for this project.

Opening the Jupyter notebooks

We will be using Jupyter notebooks to demonstrate some of the concepts and workflows described in the paper.

Jupyter notebooks give you an interactive development environment, and shows you your code, your code outputs (e.g. calculation results, processed data, visualizations) as well as other rich media (such as text, HTML, images, equations, even LaTeX!) together in a notebook-style environment. Jupyter notebooks are commonly used in the machine learning field.

You should have installed the packages required by Jupyter notebooks already if you followed the steps above to create the bestpractices environment.

In that case, you can start Jupyter by following these steps:

1. Navigate to the project directory (from above).
2. Open Anaconda prompt in this directory.
3. Run the following command from Anaconda prompt to start a Jupyter notebook server: jupyter notebook
4. Your web browser should open automatically and navigate to the Jupyter notebook webpage that has been created by the Jupyter notebook server. If not, you can look in the Anaconda prompt and find a line that says:
   • The Jupyter Notebook is running at: http://localhost:8888/?token=<token> (or something similar)
   • Navigate to this address using your favorite web browser to access the Jupyter notebooks
5. In the Jupyter notebook webpage, navigate to the notebooks directory and click on a Jupyter notebook to start the notebook
   • We recommend you to start with the first notebook, which provides an overview of the example ML project and the contents of the following notebooks.

Using the Jupyter notebooks

Jupyter notebooks are composed of several types of "cells", the most import types being code cells ("Code") and text cells ("Markdown"). You can edit code cells by clicking inside the cell and editing the code. To edit text cells, double click inside the cell and then edit the text. You can use Markdown-style formatting.

You can navigate a Jupyter notebook by using your mouse or your keyboard arrow keys.

Some other handy keyboard shortcuts to know:

For more information about how to use Jupyter notebooks, you can consult the official Jupyter Notebook documentation as well as the wealth of available information online.

Julia version via Pluto.jl

A julia implementation can be found in the folder pluto_notebooks. Additional instructions for setup are provided in the README file there. In general much of the same workflow has been kept in place the major difference is the use of julia equivalent packages (e.g., DataFrames.jl for Pandas).