elicit.simulations module

class elicit.simulations.Priors(ground_truth: bool, init_matrix_slice: Dict[str, Tensor] | None, trainer: Trainer, parameters: List[Parameter], network: NFDict | None, expert: ExpertDict, seed: int)

Bases: Module

Initializes the hyperparameters (i.e., trainable variables)

Parameters:

ground_truthbool

True if expert data are simulated from a given ground truth (oracle)

init_matrix_slicedict[str, tf.Tensor], optional

samples drawn from the initialization distribution to initialize the hyperparameter of the parametric prior distributions Only required for method=”parametric_prior” otherwise None.

trainerdict

specification of training settings and meta-information for workflow using elicit.elicit.trainer()

parameterslist

list of model parameters specified with elicit.elicit.parameter().

networkdict, optional

specification of neural network using a method implemented in elicit.networks. Only required for deep_prior method. For parametric_prior use None.

expertdict

provide input data from expert or simulate data from oracle with either the data or simulator method of the elicit.elicit.Expert module.

seedint

seed used for learning.

__init__(ground_truth: bool, init_matrix_slice: Dict[str, Tensor] | None, trainer: Trainer, parameters: List[Parameter], network: NFDict | None, expert: ExpertDict, seed: int)

__call__() → Tensor

Samples from the initialized prior distribution(s).

Returns:

prior_samplestf.Tensor, shape: [B,num_samples,num_params]

Samples from prior distribution(s).

elicit.simulations.intialize_priors(init_matrix_slice: Dict[str, Tensor] | None, method: str, seed: int, parameters: List[Parameter], network: NFDict | None) → Dict[str, Tensor]

Initialize prior distributions.

Parameters:

init_matrix_slicetf.Tensor, optional

samples drawn from the initialization distribution to initialize the hyperparameter of the parametric prior distributions Only for method=”parametric_prior”, otherwise None.

methodstr

parametric_prior or deep_prior method as specified in elicit.elicit.trainer()

seedint

seed of current workflow run as specified in elicit.elicit.trainer()

parameterslist[dict]

list of model parameter specifications using elicit.elicit.parameter().

networkdict, optional

specification of neural network using a method implemented in elicit.networks. Only required for deep_prior method. For parametric_prior use None.

Returns:

init_priordict[str, tf.Tensor]

returns initialized prior distributions ready for prior sampling.

elicit.simulations.sample_from_priors(initialized_priors: Dict[str, Variable] | Callable, ground_truth: bool, num_samples: int, B: int, seed: int, method: str, parameters: List[Parameter], network: NFDict | None, expert: ExpertDict) → Tensor

Samples from initialized prior distributions.

Parameters:

initialized_priorsdict[str, tf.Variable] or Callable

initialized prior distributions ready for prior sampling.

ground_truthbool

True if expert data is simulated from ground truth.

num_samplesint

number of samples from the prior(s).

Bint

batch size.

seedint

seed used for learning.

methodstr

parametric_prior or deep_prior method as specified in elicit.elicit.trainer()

parameterslist

list of model parameters specified with elicit.elicit.parameter().

networkdict, optional

specification of neural network using a method implemented in elicit.networks. Only required for deep_prior method. For parametric_prior use None.

expertcallable

provide input data from expert or simulate data from oracle with either the data or simulator method of the elicit.elicit.Expert module.

Returns:

prior_samplestf.Tensor, shape: [B, num_samples, num_params]

Samples from prior distributions.

elicit.simulations.simulate_from_generator(prior_samples: Tensor, seed: int, model: dict) → Dict[str, Tensor]

Simulates data from the specified generative model.

Parameters:

prior_samplestf.Tensor, shape: [B, num_samples, num_params]

samples from prior distributions.

seedint

seed used for learning. Specification in elicit.elicit.trainer().

modeldict

specification of generative model using elicit.elicit.model().

Returns:

model_simulationsdict[str, tf.Tensor]

simulated data from generative model.