elicit.initialization module

elicit.initialization.uniform_samples(seed: int, hyppar: List[str], n_samples: int, method: str, mean: float | Iterable[float], radius: float | Iterable[float], parameters: List[Parameter])

Sample from uniform distribution for each hyperparameter.

Parameters:

seedint

user-specified seed defined in elicit.elicit.trainer().

hypparlist

list of hyperparameter names (strings) declaring the order for the list of means and radius. If means and radius are each a float, then this number is applied to all hyperparameter such that no order of hyperparameter needs to be specified. In this case hyppar=None

n_samplesint

number of samples from the uniform distribution for each hyperparameter.

methodstr

name of sampling method used for drawing samples from uniform. Currently implemented are “random”, “lhs”, and “sobol”.

meanfloat or list

Specification of the uniform distribution. The uniform distribution ranges from (mean-radius) to (mean+radius).

radiusfloat or list

Specification of the uniform distribution. The uniform distribution ranges from (mean-radius) to (mean+radius).

parameterslist

list including dictionary with all information about the (hyper-)parameters. Can be retrieved as attribute from the initialized elicit.elicit.Elicit obj (i.e., eliobj.parameters)

Returns:

res_dictdict

dictionary with keys being the hyperparameters and values the samples from the uniform distribution.

Raises:

ValueError

method must be either “sobol”, “lhs”, or “random”. n_samples must be a positive integer

TypeError

arises if method is not a string.

elicit.initialization.init_runs(expert_elicited_statistics: Dict[str, Tensor], initializer: Initializer, parameters: List[Parameter], trainer: Trainer, model: Dict[str, Any], targets: List[Target], network: Optional, expert: ExpertDict, seed: int)

Computes the discrepancy between expert data and simulated data for multiple hyperparameter initialization values.

Parameters:

expert_elicited_statisticsdict

user-specified expert data as provided by elicit.elicit.Expert().

initializerdict

user-input from elicit.elicit.initializer().

parameterslist

user-input from elicit.elicit.parameter().

trainerdict

user-input from elicit.elicit.trainer().

modeldict

user-input from elicit.elicit.model().

targetslist

user-input from elicit.elicit.target().

networkdict, optional

user-input from one of the methods implemented in the elicit.networks module.

expertdict

user-input from elicit.elicit.Expert().

seedint

internal seed for reproducible results

Returns:

loss_listlist

list with all losses computed for each initialization run.

init_var_listlist

list with initializer prior model for each run.

init_matrixdict

dictionary with keys being the hyperparameter names and values being the drawn initial values per run.

elicit.initialization.init_prior(expert_elicited_statistics: Dict[str, Tensor], initializer: Initializer | None, parameters: List[Parameter], trainer: Trainer, model: Dict[str, Any], targets: List[Target], network: NFDict | None, expert: ExpertDict, seed: int)

Extracts target loss, corresponding initial hyperparameter values, and initialized prior model from init_runs().

Parameters:

expert_elicited_statisticsdict

user-specified expert data as provided by elicit.elicit.Expert().

initializerdict, optional

user-input from elicit.elicit.initializer().

parameterslist

user-input from elicit.elicit.parameter().

trainerdict

user-input from elicit.elicit.trainer().

modeldict

user-input from elicit.elicit.model().

targetslist

user-input from elicit.elicit.target().

networkdict, optional

user-input from one of the methods implemented in the elicit.networks module.

expertdict

user-input from elicit.elicit.Expert().

seedint

internally used seed for reproducible results

Returns:

init_prior_modelelicit.simulations.Priors object

initialized priors that will be used for the training phase.

loss_listlist

list with all losses computed for each initialization run.

init_priorlist

list with initializer prior model for each run.

init_matrixdict

dictionary with keys being the hyperparameter names and values being the drawn initial values per run.

elicit.initialization.uniform(radius: float | List[float] = 1.0, mean: float | List[float] = 0.0, hyper: List[str] | None = None) → Uniform

Specification of uniform distribution used for drawing initial values for each hyperparameter. Initial values are drawn from a uniform distribution ranging from mean-radius to mean+radius.

Parameters:

radiusfloat or list[float]

Initial values are drawn from a uniform distribution ranging from mean-radius to mean+radius. If a float is provided the same setting will be used for all hyperparameters. If different settings per hyperparameter are required, a list of length equal to the number of hyperparameters should be provided. The order of values should be equivalent to the order of hyperparameter names provided in hyper. The default is 1..

meanfloat or list[float]

Initial values are drawn from a uniform distribution ranging from mean-radius to mean+radius. If a float is provided the same setting will be used for all hyperparameters. If different settings per hyperparameter are required, a list of length equal to the number of hyperparameters should be provided. The order of values should be equivalent to the order of hyperparameter names provided in hyper. The default is 0..

hyperlist[str], optional

List of hyperparameter names as specified in elicit.elicit.hyper(). The values provided in radius and mean should follow the order of hyperparameters indicated in this list. If a float is passed to radius and mean this argument is not necessary.

Returns:

init_dictdict

Dictionary with all seetings of the uniform distribution used for initializing the hyperparameter values.

Raises:

AssertionError

hyper, mean, and radius must have the same length.