sampling

Random draws for PET: which stream they come from, and the sampler that makes them.

Every draw PET makes -- prior realisations, perturbed observations, outlier and crash replacement, the auto-adaptive localization's shuffle, popt's control perturbations -- goes through the ensemble's rng. That is a numpy.random.RandomState seeded from the ensemble config's seed when one is given, so a run is reproducible on its own and leaves NumPy's global state untouched; without a seed it is :class:GlobalRandomStream, which draws from the global functions exactly as PET always did, so np.random.seed(...) keeps controlling a run.

GlobalRandomStream

NumPy's global random functions behind a RandomState-shaped object.

Exists for two reasons: the numpy.random module itself cannot be pickled, and the ensemble is pickled by its emergency dump; and a named object makes it visible in code that a draw comes from the global stream.

choice(a, size=None, replace=True, p=None)

As numpy.random.choice, on the global stream.

get_state()

As numpy.random.get_state, on the global stream.

multivariate_normal(mean, cov, size=None)

As numpy.random.multivariate_normal, on the global stream.

normal(loc=0.0, scale=1.0, size=None)

As numpy.random.normal, on the global stream.

permutation(x)

As numpy.random.permutation, on the global stream.

rand(*shape)

As numpy.random.rand, on the global stream.

randn(*shape)

As numpy.random.randn, on the global stream.

set_state(state)

As numpy.random.set_state, on the global stream.

standard_normal(size=None)

As numpy.random.standard_normal, on the global stream.

uniform(low=0.0, high=1.0, size=None)

As numpy.random.uniform, on the global stream.

gen_real(mean, var, number, rng=None, limits=None, return_chol=False)

Realisations of a Gaussian with the given mean and (co)variance.

Draw for draw the same as geostat.decomp.Cholesky.gen_real -- the same shapes drawn in the same order with the same arithmetic -- so runs are bit-identical to what geostat produced; only the stream is a parameter now.

Parameters:

Name	Type	Description	Default
mean	(array - like, shape(n))	Mean vector.	required
var	array - like	Variance vector (n,), covariance matrix (n, n), or a scalar when mean has one element.	required
number	int	Number of realisations.	required
rng	RandomState - like	The stream to draw from; the global one by default.	None
limits	dict	{'lower': ..., 'upper': ...} to clip the realisations to.	None
return_chol	bool	Also return the factor used: sqrt(var) for a diagonal, the upper Cholesky factor otherwise.	False

Returns:

Type	Description
ndarray, shape (n, number), and the factor when ``return_chol``.

random_stream(seed=None)

The stream a run draws from: a private RandomState if seed is given, else the global one.