afeutrill / npd_entropy

An implementation of a non-parametric differential entropy rate estimator

npd_entropy

An implementation of a non-parametric differential entropy rate estimator, developed by Andrew Feutrill and Matthew Roughan. This technique quantises data from a continuous-valued process, then utilises Shannon entropy rate estimators for a differential entropy rate estimate.

Installation

Use the package manager pip to install npd_entropy.

pip install npd_entropy

Usage

import npd_entropy

quantised_series = npd_entropy.quantise_series(series, 1) # returns a quantised version of the series, with bin size of 1
shannon_entropy_rate_estimate = npd_entropy.grassberger_estimate(quantised_series, 100) # returns the shannon entropy rate estimate, with a history of 100 data points
npd_entropy.npd_entropy(shannon_entropy_rate_estimate, 1) #returns the differential entropy rate estimate

Contributors

• Andrew Feutrill

References

Feutrill, A., & Roughan, M. (2021). NPD Entropy: A Non-Parametric Differential Entropy Rate Estimator. arXiv preprint arXiv:2105.11580.

Please cite this paper if you use this estimator.

License

MIT