bases

jaxns.samplers.bases

Module Contents

class SeedPoint

Bases: NamedTuple

U0: jaxns.internals.types.FloatArray

log_L0: jaxns.internals.types.FloatArray

T

class BaseAbstractRejectionSampler

Bases: jaxns.samplers.abc.AbstractSampler[T], Generic[T]

Helper class that provides a standard way to create an ABC using inheritance.

class BaseAbstractMarkovSampler

Bases: jaxns.samplers.abc.AbstractSampler[T], Generic[T]

A sampler that conditions off a known satisfying point, e.g. a seed point.

abstract get_sample_from_seed(key, seed_point, log_L_constraint, sampler_state)

Produce a single i.i.d. sample from the model within the log_L_constraint.

Parameters:

• key (jaxns.internals.types.PRNGKey) – PRNGkey

• seed_point (SeedPoint) – function that gets the next sample from a seed point

• log_L_constraint (jaxns.internals.types.FloatArray) – the constraint to sample within

• sampler_state (T) – the data pytree needed and produced by the sampler

Returns:

an i.i.d. sample, and batched phantom samples

Return type:

Tuple[jaxns.nested_samplers.common.types.Sample, jaxns.nested_samplers.common.types.Sample]

abstract get_seed_point(key, sampler_state, log_L_constraint)

Samples a seed point from the live points.

Parameters:

• key (jaxns.internals.types.PRNGKey) – PRNGKey

• sampler_state (T) – the current sampler state

• log_L_constraint (jaxns.internals.types.FloatArray) – a log-L constraint to sample within. Must always be at least one sample in front above this to avoid infinite loop.

Returns:

a seed point

Return type:

SeedPoint