Weight Calculation for Semi-Supervised Learning

This package provides tools for calculating weights for test data based on training data using nearest neighbors and various distance measures. The implementation leverages scikit-learn, scipy, and joblib for efficient computation.

https://pypi.org/project/db-weights/

Features

• Calculate weights using Nearest Neighbors
• Calculate weights using different distance measures
• Supports various distance metrics including Euclidean, Mahalanobis, Cosine, etc.
• Parallel processing for faster computation

Installation

You can install the package using pip:

pip install db-weights

Usage

Importing the package

import numpy as np
from db_weights import WeightCalculator

Creating a WeightCalculator instance

weight_calculator = WeightCalculator(n_neighbors=3, algorithm='auto', n_jobs=-1)

Calculating weights using Nearest Neighbors

x_train = np.random.rand(100, 5)  # Training data
x_test = np.random.rand(20, 5)    # Test data

weights_nn = weight_calculator.calculate_weights_nn(x_train, x_test, weight=1)
print(weights_nn)

Calculating weights using distance measures

weights_dist = weight_calculator.calculate_weights_dist(x_train, x_test, weight=1, measure_type='euclidean')
print(weights_dist)

Getting available distance measures

measurements = weight_calculator.get_measurements()
print(measurements)

Make WHL

pip install setuptools wheel
python setup.py sdist bdist_wheel

Dependencies

• numpy
• scikit-learn
• scipy
• joblib

Contributing

Contributions are welcome. Please fork the repository and submit a pull request.

License

This project is licensed under the MIT License - see the LICENSE file for details.