BdeRegressor#

class bde.BdeRegressor(n_members: int = 2, hidden_layers: list[int] | None = None, seed: int = 0, loss: BaseLoss | None = None, activation: str = 'relu', epochs: int = 20, patience: int | None = None, validation_split: float = 0.15, n_samples: int = 10, warmup_steps: int = 50, lr: float = 0.001, weight_decay: float = 0.0001, n_thinning: int = 2, desired_energy_var_start: float = 0.5, desired_energy_var_end: float = 0.1, step_size_init: float | None = None, prior_family: str | PriorDist = 'standardnormal', prior_kwargs: dict[str, Any] | None = None)#

Regression-friendly wrapper exposing scikit-learn style API.

Methods

`fit`(x, y)	Fit the regression ensemble on the provided dataset.
`get_metadata_routing`()	Get metadata routing of this object.
`get_params`([deep])	Get parameters for this estimator.
`history`()	Return training history from the builder.
`predict`(x[, mean_and_std, ...])	Predict regression targets with optional uncertainty summaries.
`score`(X, y[, sample_weight])	Return coefficient of determination on test data.
`set_fit_request`(*[, x])	Configure whether metadata should be requested to be passed to the `fit` method.
`set_params`(**params)	Set the parameters of this estimator.
`set_predict_request`(*[, credible_intervals, ...])	Configure whether metadata should be requested to be passed to the `predict` method.
`set_score_request`(*[, sample_weight])	Configure whether metadata should be requested to be passed to the `score` method.

__init__(n_members: int = 2, hidden_layers: list[int] | None = None, seed: int = 0, loss: BaseLoss | None = None, activation: str = 'relu', epochs: int = 20, patience: int | None = None, validation_split: float = 0.15, n_samples: int = 10, warmup_steps: int = 50, lr: float = 0.001, weight_decay: float = 0.0001, n_thinning: int = 2, desired_energy_var_start: float = 0.5, desired_energy_var_end: float = 0.1, step_size_init: float | None = None, prior_family: str | PriorDist = 'standardnormal', prior_kwargs: dict[str, Any] | None = None)#

Initialise the regressor with architecture, optimisation, and sampling settings.

Parameters:

n_membersint, default=2: Number of deterministic networks in the ensemble.
hidden_layerslist[int] | None, default= None: Hidden layer widths; defaults to [4, 4] internally when None.
seedint, default=0: Shared PRNG seed for member initialisation and sampling.
lossBaseLoss | None: Custom training loss; defaults to bde.loss.GaussianNLL.
activationstr, default=’relu’: Activation function applied to each hidden layer.
epochsint, default=20: Maximum training epochs during the deterministic phase.
patienceint | None, optional: Early-stopping patience measured in epochs; None disables it.
validation_splitfloat, default=0.15: Fraction of data reserved for validation when early stopping is enabled. Must lie in (0, 1); when None, all data is used and early stopping is skipped.
n_samplesint, default=10: Posterior samples retained for each ensemble member.
warmup_stepsint, default=50: Number of warm-up iterations for the MCMC sampler.
lrfloat, default=1e-3: Learning rate for the Adam optimiser used in pre-sampling training.
weight_decay: float: Weight decay parameter for the AdamW optimiser applied during member training.
n_thinningint, default=2: Thinning interval applied to posterior samples.
desired_energy_var_startfloat, default=0.5: Target energy variance at the start of warm-up.
desired_energy_var_endfloat, default=0.1: Target energy variance at the end of warm-up.
step_size_initfloat | None, optional: Override for the sampler’s initial step size; falls back to lr.
prior_familystr or PriorDist: Prior distribution for network weights; accepts a PriorDist enum or string key. Defaults to "standardnormal".
prior_kwargsdict[str, Any] | None: Optional keyword arguments forwarded to the chosen prior_family (e.g. {"scale": 0.1} for a wider or narrower Normal or Laplace).

Fit the regression ensemble on the provided dataset.

Parameters:

xArrayLike: Feature matrix of shape (n_samples, n_features).
yArrayLike: Continuous targets shaped (n_samples,) or (n_samples, 1).

Returns:

BdeRegressor: The fitted estimator instance.

get_metadata_routing()#

Get metadata routing of this object.

Please check User Guide on how the routing mechanism works.

Returns:

routingMetadataRequest: A MetadataRequest encapsulating routing information.

get_params(deep=True)#

Get parameters for this estimator.

Parameters:

deepbool, default=True: If True, will return the parameters for this estimator and contained subobjects that are estimators.

Returns:

paramsdict: Parameter names mapped to their values.

history()#

Return training history from the builder.

Returns:

dict[str, dict[str, list[float]]]: Dictionary mapping member names to their training and validation loss histories.

Predict regression targets with optional uncertainty summaries.

Parameters:

xArrayLike: Feature matrix of shape (n_samples, n_features).
mean_and_stdbool, default=False: When True, return a tuple (mean, std) where both arrays have shape (n_samples,) and std combines aleatoric and epistemic uncertainty.
credible_intervalslist[float] | None, default=None: Quantile levels in (0, 1) used to summarise the predictive distribution. When not None and mean_and_std is False, the method returns (mean, quantiles) where mean has shape (n_samples,) and quantiles has shape (Q, n_samples) with one row per requested level. Note that the values are quantiles, not closed intervals; for an 80% interval you would typically pass [0.1, 0.9] and interpret the two returned quantile curves as lower and upper bounds.
rawbool, default=False: When True, ignore other flags and return the raw ensemble outputs with shape (n_members, n_samples_draws, n_samples, 2), corresponding to per-member, per-draw mean and scale parameters.

Returns:

ArrayLike | tuple[jax.Array, jax.Array]

If raw is True, the raw tensor described above.
If mean_and_std is True, a tuple (mean, std).
If credible_intervals is provided, a tuple (mean, quantiles).
Otherwise, the predictive mean (n_samples,).

score(X, y, sample_weight=None)#

Return coefficient of determination on test data.

The coefficient of determination, $R^2$, is defined as $(1 - \frac{u}{v})$, where $u$ is the residual sum of squares ((y_true - y_pred)** 2).sum() and $v$ is the total sum of squares ((y_true - y_true.mean()) ** 2).sum(). The best possible score is 1.0 and it can be negative (because the model can be arbitrarily worse). A constant model that always predicts the expected value of y, disregarding the input features, would get a $R^2$ score of 0.0.

Parameters:

Xarray-like of shape (n_samples, n_features): Test samples. For some estimators this may be a precomputed kernel matrix or a list of generic objects instead with shape (n_samples, n_samples_fitted), where n_samples_fitted is the number of samples used in the fitting for the estimator.
yarray-like of shape (n_samples,) or (n_samples, n_outputs): True values for X.
sample_weightarray-like of shape (n_samples,), default=None: Sample weights.

Returns:

scorefloat: $R^2$ of self.predict(X) w.r.t. y.

Notes

The $R^2$ score used when calling score on a regressor uses multioutput='uniform_average' from version 0.23 to keep consistent with default value of r2_score(). This influences the score method of all the multioutput regressors (except for MultiOutputRegressor).

set_fit_request(*, x: bool | None | str = '$UNCHANGED$') → BdeRegressor#

Configure whether metadata should be requested to be passed to the fit method.

Note that this method is only relevant when this estimator is used as a sub-estimator within a meta-estimator and metadata routing is enabled with enable_metadata_routing=True (see sklearn.set_config()). Please check the User Guide on how the routing mechanism works.

The options for each parameter are:

True: metadata is requested, and passed to fit if provided. The request is ignored if metadata is not provided.
False: metadata is not requested and the meta-estimator will not pass it to fit.
None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.
str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains the existing request. This allows you to change the request for some parameters and not others.

Added in version 1.3.

Parameters:

xstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED: Metadata routing for x parameter in fit.

Returns:

selfobject: The updated object.

set_params(**params)#

Set the parameters of this estimator.

The method works on simple estimators as well as on nested objects (such as Pipeline). The latter have parameters of the form <component>__<parameter> so that it’s possible to update each component of a nested object.

Parameters:

**paramsdict: Estimator parameters.

Returns:

selfestimator instance: Estimator instance.

Configure whether metadata should be requested to be passed to the predict method.

The options for each parameter are:

True: metadata is requested, and passed to predict if provided. The request is ignored if metadata is not provided.
False: metadata is not requested and the meta-estimator will not pass it to predict.
None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.
str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains the existing request. This allows you to change the request for some parameters and not others.

Added in version 1.3.

Parameters:

credible_intervalsstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED: Metadata routing for credible_intervals parameter in predict.
mean_and_stdstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED: Metadata routing for mean_and_std parameter in predict.
rawstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED: Metadata routing for raw parameter in predict.
xstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED: Metadata routing for x parameter in predict.

Returns:

selfobject: The updated object.

set_score_request(*, sample_weight: bool | None | str = '$UNCHANGED$') → BdeRegressor#

Configure whether metadata should be requested to be passed to the score method.

The options for each parameter are:

True: metadata is requested, and passed to score if provided. The request is ignored if metadata is not provided.
False: metadata is not requested and the meta-estimator will not pass it to score.
None: metadata is not requested, and the meta-estimator will raise an error if the user provides it.
str: metadata should be passed to the meta-estimator with this given alias instead of the original name.

The default (sklearn.utils.metadata_routing.UNCHANGED) retains the existing request. This allows you to change the request for some parameters and not others.

Added in version 1.3.

Parameters:

sample_weightstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED: Metadata routing for sample_weight parameter in score.

Returns:

selfobject: The updated object.

Examples using `bde.BdeRegressor`#

Usage Examples

BdeRegressor#

Examples using bde.BdeRegressor#

Examples using `bde.BdeRegressor`#