DNNPosterior

class obsidian.surrogates.custom_torch.DNNPosterior(values: Tensor)

Bases: EnsemblePosterior

__init__(values: Tensor)

Parameters:

values – Values of the samples produced by this posterior as a (b) x s x q x m tensor where m is the output size of the model and s is the ensemble size.

Methods

__init__(values)
density(value)	The probability density (or mass) of the distribution.
quantile(value)	Quantile of the ensemble posterior
rsample([sample_shape])	Sample from the posterior (with gradients).
rsample_from_base_samples(sample_shape, ...)	Sample from the posterior (with gradients) using base samples.
sample([sample_shape])	Sample from the posterior without gradients.

Attributes

base_sample_shape	The base shape of the base samples expected in rsample.
batch_range	The t-batch range.
device	The torch device of the posterior.
dtype	The torch dtype of the posterior.
ensemble_size	The size of the ensemble
mean	The mean of the posterior as a (b) x n x m-dim Tensor.
variance	The variance of the posterior as a (b) x n x m-dim Tensor.
weights	The weights of the individual models in the ensemble.

quantile(value: Tensor) → Tensor

Quantile of the ensemble posterior