This repository contains the first version of HANNA that was restricted to binary mixtures. We now provide an enhanced version of HANNA that is applicable to multi-component mixtures. You can find it in our new repo.
This repository contains the implementation of our HArd-constraint Neural Network for Activity coefficient prediction (HANNA). HANNA can be used to predict activity coefficients in any binary mixture whose components can be represented as SMILES strings. You can find details on HANNA in our paper.
You can explore HANNA and other models interactively on our new website, MLPROP, without any installation.
To set up the project, follow these steps:
-
Clone the repository:
git clone https://github.com/tspecht93/HANNA.git cd HANNA -
Core packages required:
- PyTorch
- NumPy & Pandas
- Scikit-learn
- Matplotlib
- RDKit
- Transformers
- JupyterLab
- SciPy
You can use the HANNA.ipynb notebook, which provides a demonstration of how to calculate activity coefficients for a binary mixture.
__init__.py: Initialization file for the package.HANNA.py: Contains the neural network architecture of HANNA.Own_Scaler.py: Custom scaler implementation for preprocessing.Plots.py: Function for creating and exporting the plot. Raw values will be saved as csv.Utils.py: Utility functions used throughout the project.HANNA.ipynb: Jupyter notebook demonstrating the usage of HANNA.README.md: Project documentation.License.txt: Contains license information for the HANNA.
This project is licensed under the MIT License. See the LICENSE file for details.