wangvei / elastic_tensor_ml

Machine learning models for predicting the bulk modulus of materials

elastic_tensor_ML

Machine learning models for predicting the bulk modulus of materials

This repository is a collection of notebooks exploring various machine learning (ML) models for predicting the bulk modulus (K) of materials. The dataset comes from the Materials Project, where the elastic tensors were calculated from first principles. The goal of the project is to develop an updated ML model because the previous model was trained on ~ 1500 data points, whereas now we have ~ 8500 data points. Additionally, the previous model was difficult to retrain so having a reproducible pipeline is important moving forward.

To-do

• Find a more statistically robust method for hyperparameter tuning and comparison of different ML models, possibly nested cross validation [arXiv]
• Try batch normalization or dropout with the neural networks

References

• Matminer

• Keras

• TensorFlow

Dependencies

All dependencies can be pip installed, but be aware that standard tensorflow is not yet compatible with python 3.7 (11/7/18)

• Python (3.6)
• Pymatgen
• Matminer
• scikit-learn
• Pandas
• NumPy
• Tensorflow
• Jupyter