skada.datasets.make_shifted_datasets

skada.datasets.make_shifted_datasets(n_samples_source=100, n_samples_target=100, shift='covariate_shift', noise=None, label='binary', ratio=0.9, mean=1, sigma=0.7, gamma=2, mu_regression=None, sigma_regression=None, regression_scaling_constant=27, center=(0, 2), center_cov_shift=(0, 2), random_state=None, return_X_y=True, return_dataset=False)[source]

Generate source and shift target.

Parameters:

n_samples_sourceint, default=100: It is the total number of points among one source clusters. At the end 8*n_samples points.
n_samples_targetint, default=100: It is the total number of points among one target clusters. At the end 8*n_samples points.
shifttuple, default='covariate_shift': Choose the nature of the shift. If 'covariate_shift', use covariate shift. If 'target_shift', use target shift. If 'concept_drift', use concept drift. If 'subspace', a subspace where the classes are separable independently of the domains exists. See detailed description of each shift in [1].
noisefloat or array_like, default=None: If float, standard deviation of Gaussian noise added to the data. If array-like, each element of the sequence indicate standard deviation of Gaussian noise added to the source and target data.
labelstr, default='binary': If 'binary, generates binary class labels. If 'multiclass', generates multiclass labels. If 'regression', generates regression's y-values.
ratiofloat, default=0.9: Ratio of the number of data in class 1 selected in the target shift and the sample_selection bias
meanfloat, default=1: value of the translation in the concept drift.
sigmafloat, default=0.7: multiplicative value of the concept drift.
mu_regressionnp.array|float, default=None: Will only be used if label=='regression' should be 2x1 matrix when shift != 'subspace' should be a scalar when shift == 'subspace'
sigma_regressionnp.array|float, default=None: Will only be used if label=='regression' should be a 2x2 matrix when shift != 'subspace' should be a scalar when shift == 'subspace'
regression_scaling_constant: float, default=27: Constant by which we multiply the y-value when label=='regression'
gammafloat, default=2: Parameter of the RBF kernel.
centerarray-like of shape (1, 2), default=((0, 2)): Center of the distribution.
center_cov_shiftarray-like of shape (1, 2), default=((0, 2)): Center of the covariate-shift.
random_stateint, RandomState instance or None, default=None: Determines random number generation for dataset creation. Pass an int for reproducible output across multiple function calls.
return_X_yboolean, optional (default=True): Returns source and target dataset as a pair of (X, y) tuples (for the source and the target respectively). Otherwise returns tuple of (X, y, sample_domain) where sample_domain is a categorical label for the domain where sample is taken.
return_datasetboolean, optional (default=False): When set to True, the function returns DomainAwareDataset object.

Returns:

(X, y, sample_domain)tuple if return_X_y=True

Tuple of (data, target, sample_domain), see the description below.

dataBunch

Dictionary-like object, with the following attributes.

X: ndarray: Samples from all sources and all targets given.
yndarray: Labels from all sources and all targets.
sample_domainndarray: The integer label for domain the sample was taken from. By convention, source domains have non-negative labels, and target domain label is always < 0.
domain_namesdict: The names of domains and associated domain labels.

datasetDomainAwareDataset

Dataset object.

References

[1]

Moreno-Torres, J. G., Raeder, T., Alaiz-Rodriguez, R., Chawla, N. V., and Herrera, F. (2012). A unifying view on dataset shift in classification. Pattern recognition, 45(1):521-530.