Gaussian Process KL Divergence Calculator

This repository contains a utility for calculating the Kullback-Leibler (KL) divergence over the distribution of functions represented by Gaussian Processes (GPs). The kl_divergence_gps module provides a function discrete_kl_for_gps for this purpose. This utility can be particularly useful in Bayesian optimization, probabilistic modeling, and various machine learning tasks where understanding the divergence between distributions is important.

Installation

To use this utility, follow these steps:

1. Clone this repository to your local machine.
2. Ensure you have Python installed (version 3.6 or higher).
3. Install the required dependencies by running pip install -r requirements.txt.
4. You can then import the discrete_kl_for_gps function from kl_divergence_gps module and use it in your Python code.

Usage

Here's an example of how to use the discrete_kl_for_gps function:

from kl_divergence_gps import discrete_kl_for_gps

def main():
    gps_means = [(3.2, 3.1), (3.6, 2.4), (3.5, 4.1), (2.2, 2.5)]
    gps_stds = [(1.2, 1.4), (0.5, 0.7), (1.8, 1.9), (1.3, 1.35)]

    kl_divergence = discrete_kl_for_gps(mus_list=gps_means, sigmas_list=gps_stds, upper_limit=10, lower_limit=10)
    print(kl_divergence)

if __name__ == '__main__':
    main()

In this example, gps_means and gps_stds are lists of tuples representing the means and standard deviations of Gaussian Processes respectively. upper_limit and lower_limit define the upper and lower limits of the calculation range. The function discrete_kl_for_gps returns the KL divergence over the specified range.

Contributing

Contributions to this repository are welcome. If you have any suggestions, bug fixes, or feature implementations, please feel free to open an issue or submit a pull request.

Contact

For any questions or concerns, you can contact the repository owner Ran Emuna.

Acknowledgments

This utility is inspired by the need for efficient computation of KL divergence over Gaussian Processes, used in the papers "Deep Reinforcement Learning for Human-Like Driving Policies in Collision Avoidance Tasks of Self-Driving Cars" and "Example-guided learning of stochastic human driving policies using deep reinforcement learning", concerning with probabilistic modeling, and deep reinforcement learning.

Thank you for using this utility! We hope it proves helpful in your projects and research endeavors.