Distributions for Probabilities

[Douglas G. Moore]

Sometimes we'd like to be able to use the functionality of Inform given an known probability distribution rather than one inferred from time series data. To that end, we need a function to construct a distribution which reproduces the original probabilities up to some tolerance.

This has already been done in PyInform by @jakehanson.

Proposed API

EXPORT inform_dist *inform_dist_estimated(double *probs, size_t n, double tol);

Example Usage

double probs[3] = {0.5, 0.2, 0.3};
inform_dist *dist = inform_dist_estimated(probs, 3, 1e-6);
assert(dist->counts == 10);
assert(inform_dist_get(dist, 0) == 5);
assert(inform_dist_get(dist, 1) == 2);
assert(inform_dist_get(dist, 2) == 3);