A Python library for entropy estimation.

Estimation of the entropy of a probability distribution on the basis of a finite sample is a long-standing problem in information theory. This library aims to provide implementations of most modern entropy estimators for discrete distributions.

Observations often come in an explicitly multivariate form. For example, collections of aligned transcription factor binding sites of length $L$ are naturally represented as joint observations of $L$ (potentially dependent) random variables. Neural spike train data, too, is often discretized so that one’s data consists in a collection of binary vectors such that the $i$th observation records the presence or absence of a spike in neuron $j$ in the time interval $(t_i,t_i+Δ t)$. In such cases it is usually advantageous to take the multivariate nature of the data into account and treat the source as a joint distribution over $L$ variables rather than a univariate distribution on $\mathcal{A}^L$ outcomes.