nessai.flows.distributions

Distributions to use as the ‘base distribution’ for normalising flows.

Module Contents

Classes

MultivariateNormal

A multivariate Normal with zero mean and specified covariance.

ResampledGaussian

Wrapper for ResampledGaussian.
class nessai.flows.distributions.MultivariateNormal(shape, var=1)

Bases: glasflow.nflows.distributions.Distribution

A multivariate Normal with zero mean and specified covariance.

Parameters:
shapetuple

Shape of distribution, this is used to determine the number of dimensions.

varfloat, optional

Variance of the distribution.

class nessai.flows.distributions.ResampledGaussian(shape: int | tuple, acceptance_fn: Callable, eps: float = 0.05, truncation: int = 100, trainable: bool = False)

Bases: glasflow.distributions.ResampledGaussian

Wrapper for ResampledGaussian.

Adds methods needed in nessai.

end_iteration

Function to be called at the end of an iteration.

For LARS this updates the estimate of the normalisation constant independently of the other parameters in the flow.

finalise(n_samples: int = 10000, n_batches: int = 10) → None

Finalise the estimate of the normalisation constant.