Curate Bibtex.

examples/plot_pairwise_alignment.py

@@ -35,7 +35,7 @@
 # In the case of elastic registration it is taken as energy function the
 # Fisher-Rao distance with a penalisation term, due to the property of
 # invariance to reparameterizations of warpings functions
-# :footcite:p:`srivastava+klassen_2016_analysis_elastic`.
+# :footcite:p:`srivastava+klassen_2016_functionala`.
 #
 # .. math::
 #   E[f \circ \gamma, g] = d_{FR} (f \circ \gamma, g)

skfda/datasets/_real_datasets.py

@@ -1540,6 +1540,6 @@ def fetch_mco(
 
 if fetch_mco.__doc__ is not None:  # docstrings can be stripped off
     fetch_mco.__doc__ += _mco_descr_template.format(
-        cite=":footcite:`ruiz++_2003_cariporide`",
+        cite=":footcite:`ruiz-meana++_2003_cariporide`",
         bibliography=".. footbibliography::",
     ) + _param_descr

skfda/exploratory/depth/_depth.py

@@ -231,7 +231,7 @@ class DistanceBasedDepth(Depth[FDataGrid], BaseEstimator):
             .. math::
                 D(x) = [1 + M(x, \mu)]^{-1}.
 
-            as explained in :footcite:`serfling+zuo_2000_depth_function`.
+            as explained in :footcite:`serfling+zuo_2000_general`.
 
     Examples:
         >>> import skfda

skfda/exploratory/stats/_fisher_rao.py

@@ -73,15 +73,15 @@ def _fisher_rao_warping_mean(
 
     The karcher mean :math:`\bar \gamma` is defined as the warping that
     minimises locally the sum of Fisher-Rao squared distances
-    :footcite:`srivastava+klassen_2016_analysis_orbit`.
+    :footcite:`srivastava+klassen_2016_statistical`.
 
     .. math::
         \bar \gamma = argmin_{\gamma \in \Gamma} \sum_{i=1}^{n}
          d_{FR}^2(\gamma, \gamma_i)
 
     The computation is performed using the structure of Hilbert Sphere obtained
     after a transformation of the warpings, see
-    :footcite:`srivastava++_2011_ficher-rao_orbit`.
+    :footcite:`srivastava++_2011_registration`.
 
     Args:
         warping: Set of warpings.
@@ -211,8 +211,8 @@ def fisher_rao_karcher_mean(
     equivalence class which makes the mean of the warpings employed be the
     identity.
 
-    See :footcite:`srivastava+klassen_2016_analysis_karcher` and
-    :footcite:`srivastava++_2011_ficher-rao_karcher`.
+    See :footcite:`srivastava+klassen_2016_statistical` and
+    :footcite:`srivastava++_2011_registration`.
 
     Args:
         fdatagrid: Set of functions to compute the

skfda/exploratory/stats/_stats.py

@@ -192,7 +192,7 @@ def geometric_median(
         \sum_{i=1}^N \left \| x_i-y \right \|
 
     The geometric median in the functional case is also described in
-    :footcite:`gervini_2008_estimation`.
+    :footcite:`gervini_2008_robust`.
     Instead of the proposed algorithm, however, the current implementation
     uses the corrected Weiszfeld algorithm to compute the median.
 

skfda/exploratory/visualization/_boxplot.py

@@ -117,7 +117,7 @@ class Boxplot(FDataBoxplot):
     detected in a functional boxplot by the 1.5 times the 50% central region
     empirical rule, analogous to the rule for classical boxplots.
 
-    For more information see :footcite:ts:`sun+genton_2011_boxplots`.
+    For more information see :footcite:ts:`sun+genton_2011_functional`.
 
     Args:
         fdatagrid: Object containing the data.
@@ -538,7 +538,7 @@ class SurfaceBoxplot(FDataBoxplot):
     :ref:`depth measure <depth-measures>`
     for functional data, it represents the envelope of the
     50% central region, the median curve, and the maximum non-outlying
-    envelope :footcite:`sun+genton_2011_boxplots`.
+    envelope :footcite:`sun+genton_2011_functional`.
 
     Args:
         fdatagrid: Object containing the data.

skfda/exploratory/visualization/_magnitude_shape_plot.py

@@ -41,7 +41,7 @@ class MagnitudeShapePlot(BasePlot):
     The outliers are detected using an instance of
     :class:`MSPlotOutlierDetector`.
 
-    For more information see :footcite:ts:`dai+genton_2018_visualization`.
+    For more information see :footcite:ts:`dai+genton_2018_multivariate`.
 
     Args:
         fdata: Object containing the data.

skfda/inference/hotelling/_hotelling.py

@@ -46,7 +46,7 @@ def hotelling_t2(
     the discrete representation, depending on the input.
 
     This statistic is defined in Pini, Stamm and Vantini
-    :footcite:`pini+stamm+vantini_2018_hotellings`.
+    :footcite:`pini++_2018_hotelling`.
 
     Args:
         fd1: Object with the first sample.
@@ -167,7 +167,7 @@ def hotelling_test_ind(
     tested are generated randomly.
 
     This procedure is from Pini, Stamm and Vantinni
-    :footcite:`pini+stamm+vantini_2018_hotellings`.
+    :footcite:`pini++_2018_hotelling`.
 
     Args:
         fd1: First sample of data.

skfda/misc/hat_matrix.py

@@ -149,12 +149,12 @@ class NadarayaWatsonHatMatrix(HatMatrix):
     For smoothing, :math:`\{x_1, ..., x_n\}` are the points with known value
     and :math:`\{x_1', ..., x_m'\}` are the points for which it is desired to
     estimate the smoothed value. The distance :math:`d` is the absolute value
-    function :footcite:`wasserman_2006_nonparametric_nw`.
+    function :footcite:`wasserman_2006_nonparametric`.
 
     For regression, :math:`\{x_1, ..., x_n\}` is the functional data and
     :math:`\{x_1', ..., x_m'\}` are the functions for which it is desired to
     estimate the scalar value. Here, :math:`d` is some functional distance
-    :footcite:`ferraty+vieu_2006_nonparametric_nw`.
+    :footcite:`ferraty+vieu_2006_functional`.
 
     In both cases :math:`K(\cdot)` is a kernel function and :math:`h` is the
     bandwidth.
@@ -224,7 +224,7 @@ class LocalLinearRegressionHatMatrix(HatMatrix):
     where :math:`\{t_1, t_2, ..., t_n\}` are points with known value and
     :math:`\{t_1', t_2', ..., t_m'\}` are the points for which it is
     desired to estimate the smoothed value
-    :footcite:`wasserman_2006_nonparametric_llr`.
+    :footcite:`wasserman_2006_nonparametric`.
 
     For **kernel regression** algorithm:
 
@@ -248,7 +248,7 @@ class LocalLinearRegressionHatMatrix(HatMatrix):
 
     Where :math:`c_{ik}^j` is the :math:`j`-th coefficient in a truncated basis
     expansion of :math:`X_i - X'_k = \sum_{j=1}^J c_{ik}^j` and :math:`d` some
-    functional distance :footcite:`baillo+grane_2008_llr`
+    functional distance :footcite:`baillo+grane_2009_local`
 
     For both cases, :math:`K(\cdot)` is a kernel function and :math:`h` the
     bandwidth.
@@ -431,7 +431,7 @@ class KNeighborsHatMatrix(HatMatrix):
     In both cases, :math:`K(\cdot)` is a kernel function and
     :math:`h_{i}` is calculated as the distance from :math:`x_i'` to its
     ``n_neighbors``-th nearest neighbor in :math:`\{x_1, ..., x_n\}`
-    :footcite:`ferraty+vieu_2006_nonparametric_knn`.
+    :footcite:`ferraty+vieu_2006_computational`.
 
     Used with the uniform kernel, it takes the average of the closest
     ``n_neighbors`` points to a given point.

skfda/misc/metrics/_fisher_rao.py

@@ -66,7 +66,7 @@ class FisherRaoDistance():
 
     If the observations are distributions of random variables the distance will
     match with the usual Fisher-Rao distance in non-parametric form for
-    probability distributions :footcite:`srivastava++_2011_ficher-rao`.
+    probability distributions :footcite:`srivastava++_2011_registration`.
 
     If the observations are defined in a :term:`domain` different than (0,1)
     their domains are normalized to this interval with an affine
@@ -166,7 +166,7 @@ def fisher_rao_amplitude_distance(
     .. math::
         \mathcal{R}(\gamma) = \|\sqrt{\dot{\gamma}}- 1 \|_{\mathbb{L}^2}^2
 
-    See the :footcite:`srivastava+klassen_2016_analysis_amplitude` for a
+    See the :footcite:`srivastava+klassen_2016_functionala` for a
     detailed explanation.
 
     If the observations are defined in a :term:`domain` different than (0,1)
@@ -264,7 +264,7 @@ def fisher_rao_phase_distance(
 
     where :math:`\gamma_{id}` is the identity warping.
 
-    See :footcite:`srivastava+klassen_2016_analysis_phase` for a detailed
+    See :footcite:`srivastava+klassen_2016_functionala` for a detailed
     explanation.
 
     If the observations are defined in a :term:`domain` different than (0,1)
@@ -348,7 +348,7 @@ def _fisher_rao_warping_distance(
         d_{\Gamma}(\gamma_i, \gamma_j) = cos^{-1} \left ( \int_0^1
         \sqrt{\dot \gamma_i(t)\dot \gamma_j(t)}dt \right )
 
-    See :footcite:`srivastava+klassen_2016_analysis_probability` for a detailed
+    See :footcite:`srivastava+klassen_2016_functionala` for a detailed
     explanation.
 
     If the warpings are not defined in [0,1], an affine transformation is made

skfda/misc/metrics/_mahalanobis.py

@@ -23,7 +23,7 @@ class MahalanobisDistance(BaseEstimator):
 
     Class that implements functional Mahalanobis distance for both
     basis and grid representations of the data
-    :footcite:`berrendero+bueno-larraz+cuevas_2020_mahalanobis`.
+    :footcite:`berrendero++_2020_mahalanobis`.
 
     Parameters:
         n_components: Number of eigenvectors to keep from

skfda/misc/operators/_srvf.py

@@ -29,7 +29,7 @@ class SRSF(
     This representation it is used to compute the extended non-parametric
     Fisher-Rao distance between functions, wich under the SRSF representation
     becomes the usual :math:`\mathbb{L}^2` distance between functions.
-    See :footcite:`srivastava+klassen_2016_analysis_square`.
+    See :footcite:`srivastava+klassen_2016_functionala`.
 
     The inverse SRSF transform is defined as
 
@@ -133,7 +133,7 @@ def transform(self, X: FDataGrid, y: object = None) -> FDataGrid:
 
         Let :math:`f : [a,b] \rightarrow \mathbb{R}` be an absolutely
         continuous function, the SRSF transform is defined as
-        :footcite:`srivastava+klassen_2016_analysis_square`:
+        :footcite:`srivastava+klassen_2016_functionala`:
 
         .. math::
 
@@ -184,7 +184,7 @@ def inverse_transform(self, X: FDataGrid, y: None = None) -> FDataGrid:
         Compute the inverse SRSF transform.
 
         Given the srsf and the initial value the original function can be
-        obtained as :footcite:`srivastava+klassen_2016_analysis_square`:
+        obtained as :footcite:`srivastava+klassen_2016_functionala`:
 
         .. math::
             f(t) = f(a) + \int_{a}^t q(t)|q(t)|dt

skfda/ml/classification/_depth_classifiers.py

@@ -218,11 +218,12 @@ class DDGClassifier(
     BaseEstimator,
     ClassifierMixin[Input, Target],
 ):
-    r"""Generalized depth-versus-depth (DD) classifer for functional data.
+    r"""
+    Generalized depth-versus-depth (DD) classifier for functional data.
 
     This classifier builds an interface around the DDGTransfomer.
-
-    The transformer takes a list of k depths and performs the following map:
+    The transformer takes a list of k depths and performs the following map
+    :footcite:p:`cuesta-albertos++_2017_ddgclassifier`:
 
     .. math::
         \mathcal{X} &\rightarrow \mathbb{R}^G \\
@@ -231,7 +232,6 @@ class DDGClassifier(
     Where :math:`D_i^j(x)` is the depth of the point :math:`x` with respect to
     the data in the :math:`i`-th group using the :math:`j`-th depth of the
     provided list.
-
     Note that :math:`\mathcal{X}` is possibly multivariate, that is,
     :math:`\mathcal{X} = \mathcal{X}_1 \times ... \times \mathcal{X}_p`.
 
@@ -310,12 +310,8 @@ class DDGClassifier(
         :class:`~skfda.preprocessing.dim_reduction.feature_extraction._ddg_transformer`
 
     References:
-        Li, J., Cuesta-Albertos, J. A., and Liu, R. Y. (2012). DD-classifier:
-        Nonparametric classification procedure based on DD-plot. Journal of
-        the American Statistical Association, 107(498):737-753.
+        .. footbibliography::
 
-        Cuesta-Albertos, J.A., Febrero-Bande, M. and Oviedo de la Fuente, M.
-        (2017) The DDG-classifier in the functional setting. TEST, 26. 119-142.
     """
 
     def __init__(  # noqa: WPS234
@@ -514,15 +510,18 @@ def predict(self, X: Union[NDArrayFloat, pd.DataFrame]) -> Target:
 
 
 class MaximumDepthClassifier(DDGClassifier[Input, Target]):
-    """Maximum depth classifier for functional data.
+    """
+    Maximum depth classifier for functional data.
 
-    Test samples are classified to the class where they are deeper.
+    Test samples are classified to the class where they are deeper
+    :footcite:p:`ghosh+chaudhuri_2005_maximum`.
 
     Parameters:
         depth_method:
             The depth class to use when calculating the depth of a test
             sample in a class. See the documentation of the depths module
             for a list of available depths. By default it is ModifiedBandDepth.
+
     Examples:
         Firstly, we will import and split the Berkeley Growth Study dataset
 
@@ -557,8 +556,8 @@ class MaximumDepthClassifier(DDGClassifier[Input, Target]):
         :class:`~skfda.ml.classification.DDGClassifier`
 
     References:
-        Ghosh, A. K. and Chaudhuri, P. (2005b). On maximum depth and
-        related classifiers. Scandinavian Journal of Statistics, 32, 327–350.
+        .. footbibliography::
+
     """
 
     def __init__(self, depth_method: Depth[Input] | None = None) -> None:

skfda/ml/classification/_logistic_regression.py

@@ -23,7 +23,7 @@ class LogisticRegression(
 
     This class implements the sequential “greedy” algorithm
     for functional logistic regression proposed in
-    :footcite:ts:`bueno++_2021_functional`.
+    :footcite:ts:`berrendero++_2022_functional`.
 
     .. warning::
         For now, only binary classification for functional

skfda/preprocessing/dim_reduction/variable_selection/_rkvs.py

@@ -133,7 +133,7 @@ class RKHSVariableSelection(
     misclassification error of all the classification problems with the
     reduced dimensionality. For a longer discussion about the optimality and
     consistence of this method, we refer the reader to the original
-    article [1]_.
+    article :footcite:`berrendero++_2018_use`.
 
     In practice the points are selected one at a time, using
     a greedy approach, so this optimality is not always guaranteed.
@@ -189,11 +189,7 @@ class RKHSVariableSelection(
         (10000, 3)
 
     References:
-        .. [1] J. R. Berrendero, A. Cuevas, and J. L. Torrecilla, «On the Use
-               of Reproducing Kernel Hilbert Spaces in Functional
-               Classification», Journal of the American Statistical
-               Association, vol. 113, no. 523, pp. 1210-1218, jul. 2018,
-               doi: 10.1080/01621459.2017.1320287.
+        .. footbibliography::
 
     """
 

skfda/preprocessing/dim_reduction/variable_selection/maxima_hunting.py

@@ -134,7 +134,7 @@ class MaximaHunting(
 
     For a longer explanation about the method, and comparison with other
     functional variable selection methods, we refer the reader to the
-    original article :footcite:`berrendero+cuevas+torrecilla_2016_hunting`.
+    original article :footcite:`berrendero++_2016_variable`.
 
     Parameters:
         dependence_measure (callable): Dependence measure to use. By default,

skfda/preprocessing/dim_reduction/variable_selection/recursive_maxima_hunting.py

@@ -660,7 +660,7 @@ class AsymptoticIndependenceTestStop(StoppingCondition):
     Stop when the selected point is independent from the target.
 
     It uses an asymptotic test based on the chi-squared distribution described
-    in :footcite:`szekely+rizzo_2010_brownian`. The test rejects independence
+    in :footcite:`szekely+rizzo_2009_brownian`. The test rejects independence
     if
 
     .. math::
@@ -870,7 +870,7 @@ class RecursiveMaximaHunting(
     selected by :class:`MaximaHunting` alone.
 
     This method was originally described in a special case in article
-    :footcite:`torrecilla+suarez_2016_hunting`.
+    :footcite:`torrecilla+suarez_2016_feature`.
     Additional information about the usage of this method can be found in
     :doc:`/modules/preprocessing/dim_reduction/recursive_maxima_hunting`.
 

skfda/preprocessing/registration/_fisher_rao.py

@@ -60,7 +60,7 @@ class FisherRaoElasticRegistration(
     `elastic mean`, wich is the local minimum of the sum of squares of elastic
     distances. See :func:`~elastic_mean`.
 
-    In :footcite:`srivastava+klassen_2016_analysis_elastic` are described
+    In :footcite:`srivastava+klassen_2016_functionala` are described
     extensively the algorithms employed and the SRSF framework.
 
     Args:

skfda/preprocessing/registration/_landmark_registration.py

@@ -203,7 +203,7 @@ def landmark_elastic_registration_warping(
         :math:`h_i(t^*_j)=t_{ij}`.
         The registered samples can be obtained as :math:`x^*_i(t)=x_i(h_i(t))`.
 
-        See :footcite:`ramsay+silverman_2005_functional_landmark`
+        See :footcite:`ramsay+silverman_2005_registration`
         for a detailed explanation.
 
     Args:
@@ -344,7 +344,7 @@ def landmark_elastic_registration(
     The registered samples will have their features aligned, i.e.,
     :math:`x^*_i(t^*_j)=x_i(t_{ij})`.
 
-    See :footcite:`ramsay+silverman_2005_functional_landmark`
+    See :footcite:`ramsay+silverman_2005_registration`
     for a detailed explanation.
 
     Args: