onyx.util.nonparametric – The idea here is that you have an unbounded stream of discrete-space

samples. You are interested in statistics and/or models of the time series at many time scales. This module supports the idea of keeping a number of statistics (sufficient for models) taking up space proportional to the logarithm of the number of samples seen, and covering exponentially larger intervals of the past.

class onyx.util.nonparametric.NonParametricAccumulator

Bases: object

Logarithm-space accumulator for non-parametric discrete models.

>>> a = NonParametricAccumulator()
>>> a.accum_many(xrange(5))
>>> a.accum_many(xrange(10))
>>> a.epochs
[[(8, 64), (9, 81)], [(13, 85)], [(14, 54)], [(11, 31)]]

accum_many(iterable)

accum_one(datum)

combine_stats(stats_iterable)

extend()

linear_stats(datum)

zero_stats()