This repository contains a collection of mini-projects completed as part of the Data-Driven Material Science course. These projects showcase various data manipulation, machine learning, and deep learning techniques applied to real-world material science data.
The repository is organized into the following sections:
- 1-1 Data Manipulation: Techniques for handling, cleaning, and transforming material science datasets.
- 1-2 Preprocessing and Feature Engineering: Methods to prepare data for machine learning models.
- 2-1 Regression: Building regression models for predicting material properties like band gaps.
- 2-2 Binary Classification: Developing classification models for distinguishing between material types.
- 3-1 Reduction and Clustering: Dimensionality reduction and clustering for pattern discovery in materials data.
- 3-2 Active Learning: Efficiently training models with minimal labeled data.
- 4-1 Deep Learning for Regression and Classification: Leveraging deep learning architectures for material prediction.
- 4-2 Deep Learning for Image Processing: Applying CNNs to analyze material structure images.
- 4-3 Generative Models: Developing generative models for material design.
- MP5 Deep Learning: A comprehensive project combining multiple deep learning techniques.
The following datasets are included in the project:
periodic_data.csv– Periodic table data for element properties.citrination-export-band-gaps.csv– Dataset for predicting material band gaps.processed_bmg.csv– Bulk metallic glass dataset.Wolverton_ABO3_perovskites.csv– Data on perovskite structures for materials discovery.
- Clone the repository:
git clone https://github.com/yourusername/data-driven-material-science.git cd data-driven-material-science
Note: This code is developed for educational and learning purposes as part of the Data-Driven Material Science course. The code is not intended for distribution or commercial use. If you wish to utilize it beyond personal learning, please contact me.
- Visualizations of material property trends
- Model performance metrics
- Insights into material design strategies
- Demonstrations of generative models for new material predictions
- Expanding datasets for broader material types
- Enhancing model architectures for improved accuracy
- Exploring additional generative models for material discovery
- Incorporating automated hyperparameter tuning for improved model performance
This project was developed as part of the Data-Driven Material Science course. Special thanks to instructors, peers, and open-source contributors for their invaluable support.
For any questions or collaborations, feel free to reach out via email: [email protected]