biasvariance-toolkit 0.1.0

Bias–Variance decomposition toolkit for regression (MSE) and classification (0–1 loss)

BiasVariance Toolkit

A Python toolkit for bias–variance decomposition of machine learning models.

Features

• Mean Squared Error (MSE) decomposition for regression tasks.
• 0–1 Loss decomposition for classification tasks.
• Works with both PyTorch models and scikit-learn models.

Installation

Clone the repository and install locally:

pip install -e .

Usage

from biasvariance_toolkit import estimate_bias_variance_mse, estimate_bias_variance_0_1

# Example for regression (MSE)
bias, variance, total, bias_plus_var, avg_train_loss, test_loss = estimate_bias_variance_mse(...)

# Example for classification (0-1 Loss)
bias, variance, expected_loss, empirical_loss, avg_train_loss, test_loss = estimate_bias_variance_0_1(...)

License

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