combinatorixPy 1.0.4

Development of Numerical Features/Descriptors to Describe Complex Materials for Machine Learning Modeling

combinatorixPy: Mixture Descriptors Calculator

This Python project calculates an combinatorial scheme for mixture descriptors of multi-component materials. The algorithm processes input from two .csv files and computes combinatorial mixture descriptors. This code computes combinatorial mixture descriptors based on cartesian power which is cartesian product of N-copies of a single set of descriptors, while N is the number of components.

Overview

CombinatorixPy is a Python package designed to generate combinatorial mixture descriptors for multi-component materials. It processes input data from two CSV files to compute these descriptors, facilitating advanced analysis in materials science.

Features

• Descriptor Generation: Creates mixture descriptors using a combinatorial approach based on Cartesian power, which is the Cartesian product of N copies of a single set of descriptors, where N is the number of components.
• Data Processing: Handles input from CSV files containing individual descriptors and mole fraction values for each component in mixtures.
• Filtering: Supports filtering of constant and nearly constant descriptors, as well as highly correlated pairs, with adjustable thresholds.

The algorithm requires two input files:

1. Descriptors File: Contains individual descriptors for each component.
2. Mole Fraction File: Contains mole fraction values for each component in each mixture. The main function, mixture_descriptors_to_csv, processes these inputs and generates a CSV file.

Getting Started

Prerequisites

Ensure you have Python 3.x installed. You will also need pip to install the package.

Installation

You can install the package in two different ways depending on whether you want to install from the local directory or directly from PyPI:

Option 1: Install from Local Directory

1. Download the Package:

   • Clone the repository or download the ZIP file from GitHub.
   • Extract the contents of the ZIP file.

2. Install the Package:

   • Open a command-line interface (CLI).
   • Navigate to the directory containing the extracted package files.
   • Run the following command to install the package:

   pip install .
   

Option 2: Install from PyPI If you prefer to install the package directly from PyPI, you can use the following command. Make sure to replace 1.0.1 with the desired version number if needed:

pip install combinatorixPy==1.0.1

Usage

After installing the package, you can use it in your Python code. Here’s a basic example of how to use the main function:

from combinatorixPy import initialize_dask_cluster, get_result
# initialize the cluster
config = {
        'n_workers': 3,
        'threads_per_worker': 100,
        'memory_limit': '400GB',
        'timeout': 300
    }
    
    
    # Initialize the Dask client with the provided config
    cluster = initialize_dask_cluster(config)
    client = Client(cluster)

# Call the function
get_result(
    descriptors_file_path='path/to/descriptors.csv',
    mole_fraction_file_path='path/to/mole_fractions.csv',
    output_directory='path/to/output',
    constant_threshold=0.01,
    correlation_threshold=0.9,
    batch_number=100000
)

Arguments

After installation, you can use the package's functions in your Python code. The main function get_result requires six arguments: descriptors_file_path: Path to the CSV file containing individual descriptors for each component. mole_fraction_file_path: Path to the CSV file with mole fraction values for each component in each mixture. output_directory: Directory path where the resulting CSV file will be saved. constant_threshold: Threshold for filtering out constant and nearly constant descriptors. correlation_threshold: Threshold for removing highly correlated descriptor pairs. batch_number: Batch number for processing highly correlated pairs due to large correlation matrices.

After installation, you can use the package's functions in your Python code. The main function get_result requires six arguments:

Contributing

Contributions are welcome! Please open an issue or submit a pull request if you have suggestions or improvements.

License

This project is licensed under the GNU General Public License - see the LICENSE file for details.