elicit.networks module#
- elicit.networks.NF(inference_network: Callable, network_specs: dict, base_distribution: Callable) NFDict[source]#
specification of the normalizing flow used from BayesFlow library
- Parameters:
- inference_networkcallable
type of inference network as specified by bayesflow.inference_networks.
- network_specsdict
specification of normalizing flow architecture. Arguments are inherited from chosen bayesflow.inference_networks.
- base_distributioncallable
Base distribution from which should be sampled during learning. Normally the base distribution is a multivariate normal.
- Returns:
- nf_dictdict
dictionary specifying the normalizing flow settings.
- class elicit.networks.BaseNormal[source]#
Bases:
object- __call__(num_params: int) Callable[source]#
Multivariate standard normal distribution with as many dimensions as parameters in the generative model.
- Parameters:
- num_paramsint
number of model parameters. Is internally specified in
elicit.simulations.sample_from_priors()
- Returns:
- callable
tfp.distributions object.