SMML codelength for data from the binomial distribution

This is an implementation of the polynomial time algorithm for computing the SMML codelength for data from the binomial distribution.

Strict Minimum Message Length (SMML) is criterion for statistical inference based on information theory. For general statistical models computing the SMML codelength of a model is NP-hard. This function implements the dynamic programming algorithm of Farr and Wallace (see [1]) that computes the SMML codelength for the binomial distribution in polynomial time. We assume a uniform prior on the probability of success. The algorithm is equivalent to Dijkstra's shortest path algorithm. For details on the algorithm and SMML, please see [1].

Code examples:

smml_binomial(5);

[P, msglen] = smml_binomial(10);

References:

[1] Farr, G. E.; Wallace, C. S.

The Complexity of Strict Minimum Message Length Inference

Computer Journal; May 2002, Vol. 45 Issue 3, p285