Centered Dirichlet Mixture (CDM) Entropy Estimator

Centered Dirichlet Mixture (CDM) entropy is a MATLAB toolbox providing routines for the estimation, from samples, of the entropy of a distribution over binary vectors.

Specifically, CDMentropy computes the posterior mean of entropy under a mixture of Dirichlet distributions, where the common mean of each Dirichlet is selected to reflect our prior knowledge about the data (we say that the Dirichlets are centered around a base measure). CDMentropy is designed to estimate the entropy of binary distributions, in particular spike train distributions, that are sparse (have few 1’s). It will likely work well for other binary data drawn from distributions with similar structure.

You can visit the GitHub project page, view the README, or just download the code as a zip file directly.

References:

  1. Archer E, Park IM, Pillow JW (2013). Bayesian entropy estimation for binary spike train data using parametric prior knowledge. Advances in Neural Information Processing Systems (NIPS) 26.
    (selected for spotlight presentation, top 5% of submitted)
    [bibtex|poster|pdf|code]

Leave a Reply

Your email address will not be published. Required fields are marked *