prior

jaxns.framework.prior

Module Contents

exception InvalidPriorName(name=None)[source]

Bases: Exception

Raised when a prior name is already taken.

Initialize self. See help(type(self)) for accurate signature.

Parameters:: name (Optional[str]) –

class Prior(dist_or_value, name=None)[source]

Bases: jaxns.framework.bases.BaseAbstractPrior

Represents a generative prior.

Parameters:

dist_or_value (Union[tfpd, jaxns.framework.bases.BaseAbstractDistribution, jax.numpy.ndarray]) –
name (Optional[str]) –

property dist: jaxns.framework.bases.BaseAbstractDistribution[source]

Return type:: jaxns.framework.bases.BaseAbstractDistribution

property value: jax.numpy.ndarray[source]

Return type:: jax.numpy.ndarray

parametrised(random_init=False)[source]

Convert this prior into a non-Bayesian parameter, that takes a single value in the model, but still has an associated log_prob. The parameter is registered as a hk.Parameter with added _param name suffix.

Parameters:: random_init (bool) – whether to initialise the parameter randomly or at the median of the distribution.
Returns:: A singular prior.
Return type:: SingularPrior