Miscellaneous

The module best.misc contains functions that do not fit anywhere else in best.

best.misc.logsumexp(a)

Computes in a stable way the following expression for the array a:

It uses the log-sum-exp trick.

Parameters:a (1D numpy array.) – An array. Returns: as defined above. Return type:float

best.misc.multinomial_resample(p)

This functions accepts a probability distribution over . This function returns the result of sampling times from this distribution.

Parameters:p (1D numpy array) – An array of positive numbers. Returns:An array of integers tween 0 and p.shape[0] - 1. Return type:1D numpy array of int.

Here is a small example:

from best.misc import multinomial_resample
p = [0.25, 0.25, 0.25, 0.25]
multinomial_resample(p)

This should print something like:

array([0, 2, 2, 3], dtype=int32)