cyclic_boosting.smoothing package#

Submodules#

cyclic_boosting.smoothing.base module#

Base classes for smoothers

class cyclic_boosting.smoothing.base.AbstractBinSmoother[source]#

Bases: BaseEstimator, RegressorMixin

Abstract base class for smoothers acting on bins.

Please implement the methods fit and predict.

inc_fitting = False#

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

supports_pandas = False#

class cyclic_boosting.smoothing.base.SetNBinsMixin[source]#

Bases: object

Mixin class for smoothers working on bins that saves binning information in fit to be available in predict.

set_n_bins(X_for_smoother)[source]#

cyclic_boosting.smoothing.extrapolate module#

Smoothers capable of extrapolating beyond the range seen in the training

The most common example is extrapolating a target profile or factor profile into the future, i.e. the smoother can be used to extend the profile beyond the time range seen in the training. In this case, the smoother operates on the feature time.

Note

Using these components requires regular retraining, because the extrapolation originates at the end of the training interval and will become more and more uncertain over time.

don’t use

cyclic_boosting.binning.ECdfTransformer
cyclic_boosting.binning.BinNumberTransformer
cyclic_boosting.binning.FractionalBinNumberTransformer

on features these smoothers operate on, because they just assign the largest bin number to feature values beyond the upper bound seen in the training which prevents any extrapolation.

For the time feature, LinearBinner is usually sufficient, because the amount of samples per time interval usually doesn’t vary that much over time.

Extrapolation is not needed for times within some period, e.g. for the weekday, the month day or the day number in the year. Please use normal smoothers for that.

class cyclic_boosting.smoothing.extrapolate.LinearExtrapolator(epsilon=0.0001)[source]#

Bases: AbstractBinSmoother

Smoother of one-dimensional bins that applies a linear regression to fit the bin values of y (as received from the profile function in the fit) and applies a linear interpolation in predict.

The math_utils.linear_regression() is used for the regression. The uncertainties $\sigma_i$ of y are transformed to the weights $w_i = \frac{1}{(\sigma_i)^2 + \epsilon}$ needed by linear_regression() in the fit method. An epsilon value is added to avoid zero division.

Parameters:: epsilon (float) – Epsilon value to avoid zero division when transforming the uncertainties to the weights.

Required columns in the X_for_smoother passed to fit():

column 0: …
column 1: ignored
column 2: uncertainty of the average y in each bin

These are provided by the following profile functions:

BetaMeanSigmaProfile

Used column in the predict():

column 0: …

Estimated parameters

Parameters:

alpha_ (float) – axis interception
beta_ (float) – slope

Small data example

>>> np.random.seed(4)
>>> n = 5
>>> x = np.arange(n)
>>> alpha = 3.0
>>> beta = 0.5
>>> y = alpha + beta * x
>>> y
array([ 3. ,  3.5,  4. ,  4.5,  5. ])
>>> y_err = np.array([np.mean(np.random.randn(n)) for i in range(n)])
>>> y_smeared = y + y_err
>>> y_smeared
array([ 2.96598022,  3.01021291,  4.02623841,  4.91211327,  5.04914922])

>>> X = np.c_[x, np.ones(n), y_err]

>>> from cyclic_boosting.smoothing.extrapolate import LinearExtrapolator
>>> est = LinearExtrapolator()
>>> est = est.fit(X, y_smeared)

>>> from decimal import Decimal
>>> Decimal(est.alpha_).quantize(Decimal("1.000"))
Decimal('2.971')
>>> Decimal(est.beta_).quantize(Decimal("1.000"))
Decimal('0.523')

>>> x1 = np.array([-7, -5, -3, -1, 0, 1, 2, 3, 4, 5, 6, 7, 8, 9., 20.])
>>> X1 = np.c_[x1]
>>> est.predict(X1)
array([ -0.69276543,   0.354095  ,   1.40095542,   2.44781584,
         2.97124605,   3.49467626,   4.01810647,   4.54153668,
         5.06496689,   5.5883971 ,   6.11182731,   6.63525752,
         7.15868773,   7.68211794,  13.43985026])

fit(X_for_smoother, y)[source]#

predict(X)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

cyclic_boosting.smoothing.meta_smoother module#

base module for abstract smoothers

class cyclic_boosting.smoothing.meta_smoother.NormalizationRegressionTypeSmoother(smoother, reg_type)[source]#

Bases: NormalizationSmoother, RegressionTypeSmoother

Meta-smoother to constrain all values according to their RegressionType from the predict method of the subsmoother.

Parameters:

smoother (AbstractBinSmoother) – smoother used to fit and predict on the normalized data points.
reg_type (RegressionType) – defines the regression type that is used to constrain the values.
Types (Regression) –
---------------- –
discontinuous (*) – boundries in the fit or where no fit events have been seen.
interpolating (*) – boundries in the fit.
extrapolating (*) –

fit(X_for_smoother, y)[source]#

predict(X_for_smoother)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

set_predict_request(*, X_for_smoother: bool | None | str = '$UNCHANGED$') → NormalizationRegressionTypeSmoother#

Request metadata passed to the predict 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 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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in predict.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.meta_smoother.NormalizationSmoother(smoother)[source]#

Bases: AbstractBinSmoother

Meta-smoother that normalizes the values for its subestimator.

Parameters:: smoother (AbstractBinSmoother) – smoother used to fit and predict on the normalized data points.

calc_norm(X_for_smoother, y)[source]#

Calculate the weighted mean of the target y that can then be used to give the subestimator a normalized distribution.

Parameters:

X_for_smoother (numpy.ndarray) – Multidimensional array with at least three columns. For a k-dimensional feature the first k columns contain the x-values for the smoothing. The k + 1 column are the weights of these x-values, while the k + 2 column contains the uncertainties.
smoothed_y (numpy.ndarray) – Array that contains the original target values.

fit(X_for_smoother, y)[source]#

predict(X_for_smoother)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

set_predict_request(*, X_for_smoother: bool | None | str = '$UNCHANGED$') → NormalizationSmoother#

Request metadata passed to the predict 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 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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in predict.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.meta_smoother.RegressionTypeSmoother(smoother, reg_type)[source]#

Bases: AbstractBinSmoother, SetNBinsMixin

Meta-smoother to constrain all values according to their RegressionType from the predict method of the subsmoother.

Parameters:

smoother (AbstractBinSmoother) – smoother used to fit and predict on the normalized data points.
reg_type (RegressionType) – defines the regression type that is used to constrain the values.
Types (Regression) –
---------------- –
discontinuous (*) – boundries in the fit or where no fit events have been seen.
interpolating (*) – boundries in the fit.
extrapolating (*) –

apply_cut(X_for_smoother, smoothed_y)[source]#

Constrain all values in smoothed_y according to their RegressionType.

Parameters:

X_for_smoother (numpy.ndarray) – Multidimensional array with at least three columns. For a k-dimensional feature the first k columns contain the x-values for the smoothing. The k + 1 column are the weights of these x-values, while the k + 2 column contains the uncertainties.
smoothed_y (numpy.ndarray) – Array that contains the result of the subsmoother.

fit(X_for_smoother, y)[source]#

predict(X_for_smoother)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

set_predict_request(*, X_for_smoother: bool | None | str = '$UNCHANGED$') → RegressionTypeSmoother#

Request metadata passed to the predict 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 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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in predict.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.meta_smoother.SectionSmoother(split_point, smoother_lower, smoother_upper, nan_representation=nan, epsilon=0.001)[source]#

Bases: AbstractBinSmoother

Meta-smoother that splits the fitted data into two parts which are fitted by different subsmoothers.

Parameters:

split_point (float) – value of x to split the data $x_{below} <= C_{split} < x_{above}$
smoother_lower (AbstractBinSmoother) – smoother used to fit and predict on the data points below and including the split_point
smoother_upper (AbstractBinSmoother) – smoother used to fit and predict on the data points above the split_point
nan_representation (float) – optional argument to define not a number values (default = np.nan).
epsilon (float) – Floating point accuracy when comparing with the split_point. (E.g. $x_{below} <= C_{split} + /epsilon$)

fit(X_for_smoother, y)[source]#

predict(X_for_smoother)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

set_predict_request(*, X_for_smoother: bool | None | str = '$UNCHANGED$') → SectionSmoother#

Request metadata passed to the predict 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 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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in predict.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

cyclic_boosting.smoothing.multidim module#

Multidimensional smoothers

class cyclic_boosting.smoothing.multidim.BinValuesSmoother[source]#

Bases: AbstractBinSmoother, PredictingBinValueMixin

Smoother of multidimensional bins that outputs the saved bin values of y (as received from the profile function in the fit) as the prediction.

This smoother only considers the first n_dim columns of X_for_smoother passed by the profile function. These columns are supposed to contain all the coordinates of the n_dim-dimensional bin centers.

Estimated parameters

Parameters:

smoothed_y_ (numpy.ndarray (float64, shape (n_bins,))) – the bin values for y (as received from the profile function in the fit), in this case without any smoothing; a pseudobin for the missing values is not supported
n_bins_ – see PredictingBinValueMixin

>>> from cyclic_boosting import smoothing
>>> X =    np.c_[[0., 0,  1,  1],
...              [0,  1,  0,  1],
...              [1,  1,  1,  1]]  # ignored
>>> y = np.array([90, 80, 50, 40])

>>> reg = smoothing.multidim.BinValuesSmoother()
>>> assert reg.fit(X, y) is reg
>>> assert np.allclose(reg.smoothed_y_, y)
>>> X = np.c_[[1.1, 0.4, 0.0, 0.1, 2.],
...           [1.2, 1.1, 0.4, 0.4, 0.]]
>>> reg.predict(X)
array([ 40.,  80.,  90.,  90.,  nan])

elems = ['smoothed_y_', 'bin_weights_']#

fit(X_for_smoother, y)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.multidim.GroupBySmoother(est, n_dim, index_weight_col=None)[source]#

Bases: AbstractBinSmoother

Multidimensional smoother that groups on the first k-1 columns of a k dimensional feature and smoothes a clone of the specified 1-dimensional smoother on each group.

Parameters:

est (AbstractBinSmoother) – One-dimensional smoother whose clones are fitted on the grouped columns
ndim (int) – Number of dimensions of the feature.
index_weight_col (int) – Index of weight column. If specified, rows with zero weight are removed. If None, no rows are dropped.

fit(X_for_smoother, y)[source]#

property n_group_columns#

predict(X)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.multidim.GroupBySmootherCB(est, n_dim)[source]#

Bases: GroupBySmoother

GroupBySmoother for cyclic boosting. Samples with zero weights are dropped to save memory.

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.multidim.Neutralize2DMetaSmoother(est)[source]#

Bases: AbstractBinSmoother, PredictingBinValueMixin

Meta-Smoother that takes another multi-dimensional smoother which inputs are smoothed by neutralize_one_dim_influence.

This means, the influence of one-dimensional features on two-dimensional features is removed by iteratively projecting the two-dimensional factor matrix onto one dimension and substract this weighted by the uncertainties from the matrix.

In a cyclic boosting model this prevents two-dimensional features to include the effect which should be learned by the one-dimension features.

cyclic_boosting.utils.neutralize_one_dim_influence().

fit(X_for_smoother, y)[source]#: Fit the transformer to training samples.

set_fit_request(*, X_for_smoother: bool | None | str = '$UNCHANGED$') → Neutralize2DMetaSmoother#

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

set_score_request(*, sample_weight: bool | None | str = '$UNCHANGED$') → Neutralize2DMetaSmoother#

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.multidim.PredictingBinValueMixin[source]#

Bases: SetNBinsMixin

Mixin for smoothers of multidimensional bins with predict() returning the corresponding entry in the estimated parameter smoothed_y_ which must have been calculated in fit().

Please create the array smoothed_y_ in your fit() method.

Estimated parameters

Parameters:

smoothed_y_ (numpy.ndarray (float64, shape (n_bins,))) – the bin values for y (as received from the profile function in the fit) after some smoothing; a pseudobin for the missing values is not supported; the values are indexed in lexicographical ordering using the
n_bins_ (numpy.ndarray (int64, shape (n_dims + x,))) –
number of bins in each dimension; it is permitted to append additional entries to this array. They are ignored in predict() anyway.

Please use set_n_bins() to initialize this estimated parameter in your implementation of fit().

For examples, see the subclass BinValuesSmoother.

predict(X)[source]#

class cyclic_boosting.smoothing.multidim.RegularizeToOneSmoother(threshold=2.5)[source]#

Bases: RegularizeToPriorExpectationSmoother

Smoother for multidimensional bins regularizing values with uncertainties to the prior expectation 1.

For details, see the superclass RegularizeToPriorExpectationSmoother and the underlying function cyclic_boosting.utils.regularize_to_prior_expectation().

Parameters:: threshold (float) – threshold in terms of sigma. If the significance of a factor is below the threshold, the global measurement replaces the factor. Internally cyclic_boosting.utils.regularize_to_prior_expectation() is used.

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.multidim.RegularizeToPriorExpectationSmoother(prior_expectation, threshold=2.5)[source]#

Bases: AbstractBinSmoother, PredictingBinValueMixin

Smoother of multidimensional bins regularizing values with uncertainties to a prior expectation.

For details, see cyclic_boosting.utils.regularize_to_prior_expectation().

Parameters:

prior_expectation (numpy.ndarray (float64, dim=1) or float) – The prior dominate the regularized value if the uncertainties are large.
threshold (float) –
Threshold in terms of sigma. If the significance of a value:

\[\text{sig}(x_i) = \frac{x_i - \text{prior\_expectation}_i} {\text{uncertainty}_i}\]

is below the threshold, the prior expectation replaces the value.

Required columns in the X_for_smoother passed to fit():

columns 0 to``n_dim - 1``: coordinates of the bin centers (ignored here)
column n_dim: ignored
column n_dim + 1, which must be the last: uncertainty of the average y in each bin

Doctests

>>> from cyclic_boosting import smoothing
>>> y = np.array([0, 1, 2, 3])
>>> X = np.c_[
...     [0, 0, 1, 1],
...     [0, 1, 0, 1],
...     [1]*4,
...     [0.1]*4]
>>> est = smoothing.multidim.RegularizeToPriorExpectationSmoother(1.)
>>> assert est.fit(X, y) is est
>>> y_smoothed = est.predict(X[:, :2])
>>> y_smoothed
array([ 0.03175416,  1.        ,  1.96824584,  2.98431348])
>>> np.allclose(1 - np.sqrt(((y[0] - 1) / 0.1)**2 - 2.5**2) * 0.1,
...     y_smoothed[0])
True
>>> np.allclose(1 + np.sqrt(((y[-1] - 1) / 0.1)**2 - 2.5**2) * 0.1,
...     y_smoothed[-1])
True

fit(X_for_smoother, y)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.multidim.WeightedMeanSmoother(prior_prediction=None)[source]#

Bases: AbstractBinSmoother, PredictingBinValueMixin

Smoother for multidimensional bins regularizing values with uncertainties to the weighted mean.

For details see cyclic_boosting.utils.regularize_to_error_weighted_mean().

Parameters:: prior_prediction (float) – If the prior_prediction is specified, all values are regularized with it and not with the error weighted mean.

>>> from cyclic_boosting import smoothing
>>> y_for_smoother = np.array([0.9, 0.9, 0.9, 1.8, 1.8, 0.4, 0.4])
>>> X_for_smoother = np.c_[
...     [1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0],
...     [0.0, 1.0, 0.0, 1.0, 0.0, 1.0, 1.0],
...     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
...     [0.05, 0.05, 0.05, 0.05, 0.15, 0.15, 0.05]]
>>> smoother = smoothing.multidim.WeightedMeanSmoother()
>>> smoother.fit(X_for_smoother, y_for_smoother)
>>> smoother.smoothed_y_
array([ 0.90096366,  0.90096366,  0.90096366,  1.79077293,  1.723854  ,
        0.45467402,  0.40662518])

fit(X_for_smoother, y)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

cyclic_boosting.smoothing.onedim module#

One-dimensional smoothers

class cyclic_boosting.smoothing.onedim.BinValuesSmoother[source]#

Bases: AbstractBinSmoother, PredictingBinValueMixin

Smoother that does not perform any smoothing (the identity smoother so to speak).

Smoother of one-dimensional bins that outputs the saved bin values of y (as received from the profile function in the fit) as the prediction. This smoother has no requirements on X_for_smoother passed by the profile function, because it just saves the target values y and ignores X_for_smoother completely.

Estimated parameters

Parameters:: smoothed_y_ (numpy.ndarray (float64, shape (n_bins,))) – the bin values for y (as received from the profile function in the fit) after some smoothing. a pseudobin for the missing values is not supported

>>> X = np.c_[[0., 1, 2, 3],
...           [1, 1, 1, 1]]
>>> y = np.array([90, 80, 50, 40])

>>> from cyclic_boosting.smoothing.onedim import BinValuesSmoother
>>> reg = BinValuesSmoother()
>>> assert reg.fit(X, y) is reg
>>> assert np.allclose(reg.smoothed_y_, y)
>>> X = np.c_[[3.1, 0.4, 1, 2.9, 3.4, 4.5]]
>>> reg.predict(X)
array([ 40.,  90.,  80.,  40.,  40.,  nan])

elem = 'smoothed_y_'#

fit(X_for_smoother, y)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.IsotonicRegressor(increasing='auto')[source]#

Bases: AbstractBinSmoother

Wrapper which delegates to sklearn.isotonic.IsotonicRegression. Use if you expect a feature to be monotonic. An example is given in http://scikit-learn.org/stable/modules/isotonic.html .

Parameters:

increasing (boolean or string, optional, default: "auto") –

If boolean, whether or not to fit the isotonic regression with y increasing or decreasing.

The string value “auto” determines whether y should increase or decrease based on the Spearman correlation estimate’s sign.

Note

X requires the following columns.

column0 :: …
column1 :: ignored
column2 :: weights

fit(X_for_smoother, y)[source]#

predict(X)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.LSQUnivariateSpline(interior_knots, degree=3)[source]#

Bases: AbstractBinSmoother

Wrapper which delegates to scipy.interpolate.LSQUnivariateSpline.

Parameters:

degree (int) – Degree of the smoothing spline, must be 1 <= k <= 5. The number of data points must be greater than the spline degree.
interior_knots (list) – Interior knots of the spline. Must be in ascending order.

fit(X_for_smoother, y)[source]#

predict(X)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.OrthogonalPolynomialSmoother[source]#

Bases: AbstractBinSmoother

A polynomial fit that uses orthogonal polynomials as basis functions.

Ansatz, see Blobel (http://www.desy.de/~blobel/eBuch.pdf)

fit(X_for_smoother, y)[source]#

predict(X)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.PolynomialSmoother(k)[source]#

Bases: AbstractBinSmoother

Least squares polynomial fit.

Fit a polynomial p(x) = p[0] * x**deg + ... + p[deg] of degree deg to points (x, y). Returns a vector of coefficients p that minimises the squared error.

This class delegates to numpy.polyfit.

Required columns in the X_for_smoother passed to fit():

column 0: …
column 1: ignored
column 2: uncertainty of the average y in each bin

These are provided by the following profile functions:

MeanProfile
BetaMeanSigmaProfile

Parameters:: k (int) – Degree of the polynomial.

Estimated parameters

Parameters:: coefficients_ (numpy.ndarray of length k) – Array of coefficients.

fit(X_for_smoother, y)[source]#

predict(X)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.PredictingBinValueMixin[source]#

Bases: object

Mixin for smoothers of one-dimensional bins with predict() returning the corresponding entry in the estimated parameter smoothed_y_ which must have been calculated in fit().

Please create the array smoothed_y_ in your fit() method.

Estimated parameters

Parameters:: smoothed_y_ (numpy.ndarray (float64, shape (n_bins,))) – the bin values for y (as received from the profile function in the fit) after some smoothing. a pseudobin for the missing values is not supported

For examples, see the subclasses BinValuesSmoother and BinKernelSmoother.

predict(X)[source]#

class cyclic_boosting.smoothing.onedim.RegularizeToOneSmoother(threshold=2.5)[source]#

Bases: RegularizeToPriorExpectationSmoother

Smoother for one-dimensional bins regularizing values with uncertainties to the mean of the values.

Parameters:: threshold (float) – threshold in terms of sigma. If the significance of a factor is below the threshold, the global measurement replaces the factor.

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.RegularizeToPriorExpectationSmoother(prior_expectation, threshold=2.5)[source]#

Bases: AbstractBinSmoother, PredictingBinValueMixin

Smoother of one-dimensional bins regularizing values with uncertainties to a prior expectation.

Parameters:

prior_expectation (numpy.ndarray (float64, dim=1) or float) – The prior dominate the regularized value if the uncertainties are large.
threshold (float) –
Threshold in terms of sigma. If the significance of a value:

\[\text{sig}(x_i) = \frac{x_i - \text{prior\_expectation}_i} {\text{uncertainty}_i}\]

is below the threshold, the prior expectation replaces the value.

Required columns in the X_for_smoother passed to fit():

column 0: …
column 1: ignored
column 2: uncertainty of the average y in each bin

Doctests

>>> y = np.array([0, 1, 2, 3])
>>> X = np.c_[
...     [0, 1, 2, 3],
...     [1]*4,
...     [0.1]*4]
>>> from cyclic_boosting.smoothing.onedim import RegularizeToPriorExpectationSmoother
>>> est = RegularizeToPriorExpectationSmoother(1.)
>>> assert est.fit(X, y) is est
>>> y_smoothed = est.predict(X[:, [0]])
>>> y_smoothed
array([ 0.03175416,  1.        ,  1.96824584,  2.98431348])
>>> np.allclose(1 - np.sqrt(((y[0] - 1) / 0.1)**2 - 2.5**2) * 0.1,
...     y_smoothed[0])
True
>>> np.allclose(1 + np.sqrt(((y[-1] - 1) / 0.1)**2 - 2.5**2) * 0.1,
...     y_smoothed[-1])
True

fit(X_for_smoother, y)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.SeasonalLassoSmoother(min_days=300)[source]#

Bases: AbstractBinSmoother

Seasonal Smoother of one-dimensional bins that applies a first order sine/cosine fit to smooth the bin values of y and returns the smoothed values as the prediction.

Lasso Regularization is used to reduce the influence of outliers.

If the number of days in the year where data is available remains below the configured threshold, the fallback value 0 is returned.

Parameters:: min_days (int) – The smoother is only fitted if more than min_days bins of data are available. Otherwise, it is set to fallback_mode and no smoothing is done. (default=300)

Notes

Required columns in the X_for_smoother passed to fit():

column 0: …
column 1: The number of samples in the bin
column 2: uncertainty of the average y in each bin (ignored)

est#

The fitted sklearn.LassoLarsIC estimator

Type:: sklearn.estimator

fit(X_for_smoother, y)[source]#

predict(X)[source]#

prepare_X(x)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.SeasonalSmoother(offset_tozero=True, order=1, custom_fit_function=None, fallback_value=0.0)[source]#

Bases: AbstractBinSmoother

Seasonal Smoother of one-dimensional bins that applies an sinus/cosinus fit to smooth the bin values of y (as received from the profile function in the fit) and returns the polynomial values as the prediction.

By default, SeasonalSmoother uses the the function f(x) = c_const + c_sin * np.sin(x) + c_cos * np.cos(x) for the fit. The constant offset c_const is ignored if offset_tozero is set ot True (default). With higher order, terms with frequency 2*x and possibly 3*x are added.

Instead of specifying an order, a custom fit-function can be supplied via the custom_fit_function argument.

Parameters:

offset_tozero (bool) – If True sets the constant offset to zero (default=True)
order (int) – Order k of the series sin(k * x) + cos(k * x). Per default this is 1 and supported for order 2 and 3.
custom_fit_function (function) – User defined function to define a custom fit function for the seasonal smoother. In this case the order parameter is ignored.
fallback_value (float) – User defined parameter that is used when the fit does not converge

Notes

def my_custom_fit_function(x, *params):
    return ...

Required columns in the X_for_smoother passed to fit():

column 0: …
column 1: ignored
column 2: uncertainty of the average y in each bin

These are provided by the following profile functions:

BetaMeanSigmaProfile
MeanProfile

frequency_#

Frequency calculated from the maximum binnumber.

Type:: float

par_#

Parameters from the fit.

Type:: numpy.ndarray

fit(X_for_smoother, y)[source]#

predict(X)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

transform_X(X)[source]#

class cyclic_boosting.smoothing.onedim.UnivariateSplineSmoother(k=3, s=None)[source]#

Bases: AbstractBinSmoother

Smoother of one-dimensional bins that applies a univariate spline fit to smooth the bin values of y (as received from the profile function in the fit) and returns the spline values as the prediction.

This class delegates to scipy.interpolate.UnivariateSpline.

Required columns in the X_for_smoother passed to fit():

column 0: …
column 1: weight sum in each bin

These are provided by the following profile functions:

MeanProfile
BetaMeanSigmaProfile

Parameters:

k (int) – Degree of the smoothing spline, must be <= 5. The number of the data points must be larger than the spline degree.
s (float or None) –
Positive smoothing factor used to choose the number of knots. The number of knots will be increased until the smoothing condition is satisfied:

\[\sum_i w_i \cdot (y_i - s(x_i))^2 \le s\]

If None (default), s = len(w) is taken which should be a good value if 1 / w[i] is an estimate of the standard deviation of y[i]. If 0, the spline will interpolate through all data points.

See scipy.interpolate.UnivariateSpline.

Estimated parameters

Parameters:: spline_ (scipy.interpolate.UnivariateSpline) – spline function that can be applied to x values

fit(X_for_smoother, y)[source]#

predict(X)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

class cyclic_boosting.smoothing.onedim.WeightedMeanSmoother(prior_prediction=None)[source]#

Bases: AbstractBinSmoother, PredictingBinValueMixin

Smoother for one-dimensional bins regularizing values with uncertainties to the weighted mean or a user defined value.

Parameters:: prior_prediction (float) – If the prior_prediction is specified, all values are regularized with it and not with the error weighted mean.

>>> y_for_smoother = np.array([0.9, 0.9, 0.9, 1.8, 1.8, 0.4, 0.4])
>>> X_for_smoother = np.c_[
...     [1.0, 0.0, 0.0, 1.0, 1.0, 0.0, 0.0],
...     [0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0],
...     [0.05, 0.05, 0.05, 0.05, 0.15, 0.15, 0.05]]
>>> from cyclic_boosting.smoothing.onedim import WeightedMeanSmoother
>>> smoother = WeightedMeanSmoother()
>>> smoother.fit(X_for_smoother, y_for_smoother)
>>> smoother.smoothed_y_
array([ 0.90096366,  0.90096366,  0.90096366,  1.79077293,  1.723854  ,
        0.45467402,  0.40662518])

>>> y_for_smoother = np.array([-0.70710678, -0.70710678])
>>> X_for_smoother = np.c_[[0., 1.], [0.012]*2, [9.12870929]*2]
>>> smoother = WeightedMeanSmoother()
>>> smoother.fit(X_for_smoother, y_for_smoother)
>>> smoother.smoothed_y_
array([-0.70710678, -0.70710678])

>>> y_for_smoother = np.array([-1, -1, -1, 1, 1])
>>> X_for_smoother = np.c_[
...     [1.0, 1.0, 1.0, 0.0, 0.0],
...     [0.0, 0.0, 0.0, 0.0, 0.0],
...     [0.5, 0.5, 0.5, 0.5, 0.5]]
>>> smoother = WeightedMeanSmoother(prior_prediction=0)
>>> smoother.fit(X_for_smoother, y_for_smoother)
>>> smoothed_abs = (4. * 1 + 1 * 0) / (4 + 1)
>>> np.testing.assert_allclose(smoother.smoothed_y_,
...     y_for_smoother * smoothed_abs)

fit(X_for_smoother, y)[source]#

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

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.

New 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:: X_for_smoother (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for X_for_smoother parameter in fit.
Returns:: self – The updated object.
Return type:: object

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

Request metadata passed to the score 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 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.

New 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_weight (str, True, False, or None, default=sklearn.utils.metadata_routing.UNCHANGED) – Metadata routing for sample_weight parameter in score.
Returns:: self – The updated object.
Return type:: object

cyclic_boosting.smoothing.orthofit module#

cyclic_boosting.smoothing.orthofit.custom_clip(value, lower, upper)[source]#

cyclic_boosting.smoothing.orthofit.cy_apply_orthogonal_poly_fit_equidistant(binnos, parameters, n_degrees)[source]#

cyclic_boosting.smoothing.orthofit.cy_orthogonal_poly_fit_equidistant(binnos: Array(float64, 1, 'A', False, aligned=True), y_values: Array(float64, 1, 'A', False, aligned=True), y_errors: Array(float64, 1, 'A', False, aligned=True))[source]#

cyclic_boosting.smoothing.orthofit.fit_orthogonal_poly_(x: Array(float64, 1, 'A', False, aligned=True), y: Array(float64, 1, 'A', False, aligned=True), weights: Array(float64, 1, 'A', False, aligned=True), parameters: Array(float64, 1, 'A', False, aligned=True)) → float64[source]#

cyclic_boosting.smoothing.orthofit.reduce_to_signifcant_parameters(n_signi_par, parameters, n_degrees, n_supporting_points)[source]#

cyclic_boosting.smoothing.orthofit.significant_parameters_(parameters, n_degrees, n_supporting_points)[source]#

Module contents#

Smoothers of target profiles and factor profiles or for the regularization of plots.

class cyclic_boosting.smoothing.RegressionType(value)[source]#

Bases: Enum

Type of regression that is supported by the smoother.

Three variants are supported:

discontinuous: It cannot be interpolated between the values seen
by the smoother (e.g. the values are unordered labels). Therefore all values in predict are set to nan that are out of the bin boundries in the fit or where no fit events have been seen.
interpolating: Interpolation is possible between the values seen by
the smoother. So only values above or below the bin boundries are critical. Therefore all values in predict are set to nan that are out of the bin boundries in the fit.
extrapolating: Extrapolation allows arbitrary values independent of
the values seen in the fit. Therefore no restrictions apply.

discontinuous = 'discontinuous'#

extrapolating = 'extrapolating'#

interpolating = 'interpolating'#