skada.KMMReweightAdapter

class skada.KMMReweightAdapter(kernel='rbf', gamma=None, degree=3, coef0=1, B=1000.0, eps=None, tol=1e-06, max_iter=1000, smooth_weights=False, solver='frank-wolfe')

Kernel Mean Matching (KMMReweight).

The idea of KMMReweight is to find an importance estimate w(x) such that the Maximum Mean Discrepancy (MMD) divergence between the target input density p_target(x) and the reweighted source input density w(x)p_source(x) is minimized.

See [23] for details.

Parameters:

kernelstr, default="rbf"

Kernel

gammafloat, None

Parameters for the kernels.

degreeint, 3

Parameters for the kernels.

coef0float, default

Parameters for the kernels.

Bfloat, default=1000.

Weight upper bound.

epsfloat, default=None

KMMReweight tolerance parameter. If None, eps is set to (sqrt(n_samples_source) - 1) / sqrt(n_samples_source).

tolfloat, default=1e-6

Tolerance for the stopping criterion in the optimization.

max_iterint, default=100

Number of maximum iteration before stopping the optimization.

smooth_weightsbool, default=False

If True, the weights are "smoothed" using the kernel function.

solverstring, default='frank-wolfe'

Available solvers : ['frank-wolfe', 'scipy'].

References

[23]

J. Huang, A. Gretton, K. Borgwardt, B. Schölkopf and A. J. Smola. 'Correcting sample selection bias by unlabeled data.' In NIPS, 2007.

Attributes:

`source_weights_`array-like, shape (n_samples,)

The learned source weights.

`X_source_`array-like, shape (n_samples, n_features)

The source data.

fit(X, y=None, *, sample_domain=None)

Fit adaptation parameters.

Parameters:

Xarray-like, shape (n_samples, n_features)

The source data.

yarray-like, shape (n_samples,)

The source labels.

sample_domainarray-like, shape (n_samples,)

The domain labels (same as sample_domain).

Returns:

selfobject

Returns self.

set_fit_request(*, sample_domain: bool | None | str = '$UNCHANGED$') → KMMReweightAdapter

Request metadata passed to the fit method.

Note that this method is only relevant if enable_metadata_routing=True (see sklearn.set_config()). Please see 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.

Note

This method is only relevant if this estimator is used as a sub-estimator of a meta-estimator, e.g. used inside a Pipeline. Otherwise it has no effect.

Parameters:

sample_domainstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED

Metadata routing for sample_domain parameter in fit.

Returns:

selfobject

The updated object.

set_transform_request(*, allow_source: bool | None | str = '$UNCHANGED$', sample_domain: bool | None | str = '$UNCHANGED$') → KMMReweightAdapter

Request metadata passed to the transform method.

Note that this method is only relevant if enable_metadata_routing=True (see sklearn.set_config()). Please see User Guide on how the routing mechanism works.

The options for each parameter are:

• True: metadata is requested, and passed to transform 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 transform.

• 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.

Note

This method is only relevant if this estimator is used as a sub-estimator of a meta-estimator, e.g. used inside a Pipeline. Otherwise it has no effect.

Parameters:

allow_sourcestr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED

Metadata routing for allow_source parameter in transform.

sample_domainstr, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED

Metadata routing for sample_domain parameter in transform.

Returns:

selfobject

The updated object.

Examples using skada.KMMReweightAdapter

Reweighting method example on covariate shift dataset

Reweighting method example on covariate shift dataset