Repository containing the files used while completing the course project of CH5650: Molecular Data Science and Informatics offered at IIT Madras during the Jan-May 2022 semester. This project aimed at recreating the results of the paper titled "Insightful classification of crystal structures using deep learning" (DOI: 10.1038/s41467-018-05169-6) [1] and extending the work to use unsupervised learning techniques to perform crystal structure classification. Filestructure of this repository:
- The report for the project is titled "MM19B057_Report.pdf";
- The presentation given in class based on the reference paper [1] is titled "Presentation.ppt";
- The Jupyter (technically Colab) notebook to recreate the results is titled "CH5650:CNN_Classify.ipynb";
- The Jupyter (also technically Colab) notebook for the unsupervised learning part of the project is titled "CH5650:Unsupervised_Classify.ipynb";
- The trained CNN model is saved as "Trained_CNN.hdf5".
The data used for this project was taken from the original work [1], and is hosted on the following links:
Pristine dataset (Diffraction fingerprints of crystals): https://dataverse.harvard.edu/api/access/datafile/3238702?format=original
Pristine dataset (Labels of crystals): https://dataverse.harvard.edu/api/access/datafile/3238704?format=original
Dataset info: https://dataverse.harvard.edu/api/access/datafile/3238706?format=original
Defected dataset (Diffraction fingerprints of defected crystals): https://dataverse.harvard.edu/api/access/datafile/3238702?format=original
Defected dataset (Labels of defected crystals): https://dataverse.harvard.edu/api/access/datafile/3238704?format=original
References: [1] A. Ziletti, D. Kumar, M. Scheffler, and L. M. Ghiringhelli, “Insightful classification of crystal structures using deep learning”, Nature Communications, vol. 9, pp. 2775 (2018)