Machine Learning for Material properties

General Information

This project trains Machine Learning (ML) models using chemical compound data, inspired by the publication:
ElemNet

The goal is to predict the "energy above the hull" (formation energy from elemental composition) using Silicon as the primary element. This project could be extended for future elements.

Dataset Source:

The dataset is obtained via the Next-Gen Materials Project API:Materials Project (MPR). To pull dataset, download (free) api from MPR.

ML Models Used:

• Linear Regression
• Random Forest
• Deep Neural Networks (DNN)
• XGBoost

ML Frameworks:

• TensorFlow for DNN
• Scikit-Learn for Random Forest & Linear Regression

ML features:

nelements, density, energy_per_atom, formation_energy_per_atom, band_gap, cbm, vbm, vpa, magmom_pa. The last two features (vpa,magmom_pa) are computed from three original descriptors: volume, total_magnetization and nsites, the two former are normalized by the last descriptor.

DNN hyperparameters:

• Slightly modified the hyperparameters of ElemNet paper.
• Split datasets to train, validation and test.

---

Technical Details

• Python Version: Python 3.10
• Notebook: Jupyter Notebook
• Environment Manager: Conda

Setting up the Environment:

Required packages to set the environment

# Create & activate environment
conda create -n chemenv
conda activate chemenv

# Install required packages:
pip install tensorflow==2.12.0
pip install scikit-learn
pip install numpy==1.24.3 
pip install pandas
pip install matplotlib
pip install scipy
pip install mp-api
pip install xgboost
pip install seaborn

The above packages are add to this mlenv.yml file. To set up the environment run:

conda env create -f env_ml.yml

Tunning DNN hyper-parameters

Dry run the hyperparametrization code:

    --HD 256 512 1024 \
    --NHD 10 17 \
    --BS 256 512 ```

To hyperparameterize number of nodes in hidden layers(HD), number of DNN layers (NHD), and batch size (BS):

``` python dnn_hyperpar.py \
    --npzfile \path\to\npzfile \
    --HD 256 512 1024 \
    --NHD 10 17 \
    --BS 256 512 ```
---

### Todo:
- Train ML classification by categorizing stable vs unstable element based on element above hull limit