pillow lab @ princeton

neural coding & computation group

Bayesian Entropy estimator for binary vector observations - matlab code

description: Computes the Bayes' least squares estimate (i.e., posterior mean) of the entropy of a discrete distribution over binary vectors under a Centered Dirichlet Mixture (CDM) prior. The CDM prior is a mixture of Dirichlet distributions with base measure given by either a parametric Bernoulli prior (DBer) or population synchrony (DSyn) prior.

The base measure serves to regularize the estimate of the discrete probability distribution:
(1) the Bernoulli prior specifies the (independent) probability of a spike p for each neuron, which is useful when spikes are rare.
(2) The population synchrony prior specifies a distribution over the number of total spikes in a word (P(0 spikes), P(1 spike), P(2 spikes),...).

download: master.zip   (more info: README.html)
browse: github project page

relevant publication:
  • Archer E, Park I, & Pillow JW (2013). Bayesian entropy estimation for binary spike train data using parametric prior knowledge. Advances in Neural Information Processing Systems 26, 1700-1708. [abs | pdf ]

  • Please report any bugs to pillow@princeton.edu