https://nanohub.org/tools/mrsicmsnotes/

1. nanoHUB jupyter tools are run on a READ ONLY filesystem. Do not perform actual disk IO from a published notebook.
2. one day I’ll turn this into a dev guide for nanohub….

clone the repository and run:

$ conda env create -f mrsicms.yaml

set up a notebook server and kernel as usual and proceed.

This is an abridged, tutorial version of the model development performed for publication at (citation to be generated)

These notebooks are available only as ipython notebooks. The complete notebook are also available in org markup language.

1. navigate to https://nanohub.org/tools/jupyter70/
2. start a terminal and clone the repository.
3. use the nanoHUB jupyter browser to start the mrsicms kernel from the “new” drop-down menu on the right.
4. develop in a notebook online

in order to access the kernel through an interactive shell:

1. start an in-browser terminal (does not work through ssh)
2. $ use -e anaconda-7
3. $ jupyter console –existing

These are just notes on what worked for me once. This should not be necessary to do.

1. conda install -c intel intel-aikit
2. conda update scipy
3. conda install -c intel mkl=2021.4.0
4. pip install git+https://github.com/PanayotisManganaris/cmcl.git
5. pip install git+https://github.com/PanayotisManganaris/yogi.git
6. pip install git+https://github.com/PanayotisManganaris/spyglass.git
7. conda install -c intel tensorboard

Methods rooted in data science, machine learning, and artificial intelligence have become necessary components of materials design endeavors, being frequently applied in conjunction with both computational and experimental data. The now thriving field of materials informatics has seen the accelerated discovery of new battery materials, solar cell absorbers, thermoelectrics, and routes for autonomous synthesis and characterization. The need to educate and train the materials science workforce on the essential elements of machine learning has never been greater. This proposal aims to establish a recurring series of tutorials at MRS spring and fall meetings that introduce newcomers to all the basic concepts of machine learning in materials science, walking them through a few interactive examples that use existing datasets and ML resources.

In this tutorial, there will be short overview presentations of several key ML concepts, following which the presenter and audience will together work through Python notebooks that contain easy-to-follow examples from the literature. The audience will likely constitute undergraduate and graduate students looking to get started with materials informatics, but the tutorial will be welcome and useful for any researcher. Some familiarity with writing code and making plots in Python would be useful.

Prerequisites: Basic familiarity with Python and some data science, basic background do materials science and engineering / materials design.

Arun

high-level examples of ML in materials science

Arun

read database, compute descriptors, csv file, descriptors, train RFR regression model, nuts and bolts

Saaketh

deep learning with a few examples

Saaketh

CNN for classification

Gilad or Arun+Saaketh

Bayesian optimization / autonomous experiments / Walkthrough using a simple example

These notebooks are the first in a series of tutorials planned for recurring workshops hosted at the MRS spring and fall meetings. It aims to introduces newcomers to an example of rigorous model engineering. This is done by interactively guiding users through the task of creating models of semiconductor band gaps using a subset of the Mannodi Research Group’s computational cubic Perovskites dataset.

Mannodi-Kanakkithodi, A., & Chan, M. K. Y. (2021). Data-driven design of novel halide perovskite alloys. Energy and Environmental Science, (), . http://dx.doi.org/10.1039/D1EE02971A