smac.epm.multi_objective_epm

Classes

MultiObjectiveEPM(target_names, configspace, ...)

Wrapper for the surrogate models to predict multiple targets.

class smac.epm.multi_objective_epm.MultiObjectiveEPM(target_names, configspace, types, bounds, seed, instance_features=None, pca_components=None, model_kwargs=None)

Bases: smac.epm.base_epm.BaseEPM

Wrapper for the surrogate models to predict multiple targets.

Only a list with the target names and the types array for the underlying model are mandatory. All other hyperparameters to model can be passed via kwargs. Consult the documentation of the corresponding model for the hyperparameters and their meanings.

Parameters

• target_names (list) – List of str, each entry is the name of one target dimension. Length of the list will be n_objectives.

• types (List[int]) – Specifies the number of categorical values of an input dimension where the i-th entry corresponds to the i-th input dimension. Let’s say we have 2 dimension where the first dimension consists of 3 different categorical choices and the second dimension is continuous than we have to pass [3, 0]. Note that we count starting from 0.

• bounds (List[Tuple[float, float]]) – bounds of input dimensions: (lower, uppper) for continuous dims; (n_cat, np.nan) for categorical dims

• instance_features (np.ndarray (I, K)) – Contains the K dimensional instance features of I different instances

• pca_components (float) – Number of components to keep when using PCA to reduce dimensionality of instance features. Requires to set n_feats (> pca_dims).

• model_kwargs (Optional[Dict[str, Any]]:) – arguments for initialing estimators

target_names

target names

num_targets

number of targets

Type

int

estimators

a list of estimators predicting different target values

Type

List[BaseEPM]

abstract construct_estimators(configspace, types, bounds, model_kwargs)

Construct a list of estimators. The number of the estimators equals ‘self.num_targets’ :type configspace: ConfigurationSpace :param configspace: Configuration space to tune for. :type configspace: ConfigurationSpace :type types: List[int] :param types: Specifies the number of categorical values of an input dimension where

the i-th entry corresponds to the i-th input dimension. Let’s say we have 2 dimension where the first dimension consists of 3 different categorical choices and the second dimension is continuous than we have to pass [3, 0]. Note that we count starting from 0.

Parameters

• bounds (List[Tuple[float, float]]) – bounds of input dimensions: (lower, uppper) for continuous dims; (n_cat, np.nan) for categorical dims

• model_kwargs (Dict[str, Any]) – model kwargs for initializing models

Returns

estimators – A list of estimators

Return type

List[BaseEPM]

predict_marginalized_over_instances(X)

Predict mean and variance marginalized over all instances.

Returns the predictive mean and variance marginalised over all instances for a set of configurations.

Parameters

X (np.ndarray of shape = [n_features (config), ]) –

Return type

Tuple[ndarray, ndarray]

Returns

• means (np.ndarray of shape = [n_samples, n_objectives]) – Predictive mean

• vars (np.ndarray of shape = [n_samples, n_objectives]) – Predictive variance

previous

smac.epm.gaussian_process

next

smac.epm.random_epm