This repository is meant to reports existing models in machine learning for materials informatics.
Python scripts are mainly adapted from Jupyter Notebook files taken from the course of Prof. M. Buehler (MIT) on Materials Informatics. You may want to follow this course to have an excellent overview of techniques used in materials informatics (https://professional.mit.edu/course-catalog/machine-learning-materials-informatics).
Here, I provide python script versions for those who are not comfortable with Jupyter Notebook and prefer working with basic scirpts.
Note that all python scripts are optimized for being visualized with VI on 100 columns preferentially.
- 0_Various_Autoencoder_leading_to_CNN.py
The above file provides various examples of autoencoders
- 1_Image_data_augmentation_example.py
The above file provide an example for data augmentation using Keras with preprocessing layers
See also the following links for data augmentation:
https://medium.com/featurepreneur/data-augmentation-using-keras-preprocessing-layers-6cdc7d49328e https://machinelearningmastery.com/image-augmentation-deep-learning-keras/
GNN with pytorch https://www.blopig.com/blog/2022/02/how-to-turn-a-smiles-string-into-a-molecular-graph-for-pytorch-geometric/