[ENH] all regressors to use "dist" kwarg for distribution input

skpro/regression/cyclic_boosting.py

@@ -83,11 +83,10 @@ class CyclicBoosting(BaseProbaRegressor):
         be on a bounded interval, with support between ``lower`` and ``upper``.
     maximal_iterations : int, default=10
         maximum number of iterations for the cyclic boosting algorithm
-    dist_type: str, one of ``'normal'`` (default), ``'logistic'``
-        inner base distribution to use for the Johnson QPD, i.e., before
-        arcosh and similar transformations.
-        Available options are ``'normal'`` (default), ``'logistic'``,
-        or ``'sinhlogistic'``.
+    dist: str, default='normal',
+        One of ``'normal'`` or ``'logistic'``inner base distribution to use for
+        the Johnson QPD, i.e., before arcosh and similar transformations. Available
+        options are ``'normal'`` (default), ``'logistic'`` or ``'sinhlogistic'``.
 
     Attributes
     ----------
@@ -133,6 +132,11 @@ class CyclicBoosting(BaseProbaRegressor):
         "tests:vm": True,  # requires its own test VM to run
     }
 
+    # TODO (release 2.14.0)
+    # remove the 'dist_type' argument from '__init__' signature
+    # remove the following 'if' check and deprecation warning
+    # de-indent the following 'else' check
+
     def __init__(
         self,
         feature_groups: Union[List[str], List[Tuple[str, ...]], None] = None,
@@ -142,7 +146,8 @@ def __init__(
         lower: Union[float, None] = None,
         upper: Union[float, None] = None,
         maximal_iterations=10,
-        dist_type: Union[str, None] = "normal",
+        dist_type: Union[str, None] = "deprecated",
+        dist: Union[str, None] = "normal",
         dist_shape: Union[float, None] = 0.0,
     ):
         self.feature_groups = feature_groups
@@ -153,10 +158,28 @@ def __init__(
         self.upper = upper
         self.maximal_iterations = maximal_iterations
         self.dist_type = dist_type
+        self.dist = dist
         self.dist_shape = dist_shape
 
         super().__init__()
 
+        # handle deprecation of dist_type -> dist
+        if dist_type != "deprecated":
+            from warnings import warn
+
+            warn(
+                "in `CyclicBoosting`, parameter 'dist_type' "
+                "will be renamed to 'dist' in version 2.14.0. "
+                "To keep current behaviour and to silence this warning, "
+                "use 'dist' instead of 'dist_type', "
+                "set dist explicitly via kwarg, and do not set dist_type.",
+                category=DeprecationWarning,
+                stacklevel=2,
+            )
+            self._dist = dist_type
+        else:
+            self._dist = dist
+
         self.quantiles = [self.alpha, 0.5, 1 - self.alpha]
         self.quantile_values = list()
         self.quantile_est = list()
@@ -315,7 +338,7 @@ def _predict_proba(self, X):
             "qv_high": self.quantile_values[2].reshape(-1, 1),
             "lower": self.lower,
             "upper": self.upper,
-            "base_dist": self.dist_type,
+            "base_dist": self._dist,
             "index": index,
             "columns": y_cols,
         }

skpro/regression/gam/_gam.py

@@ -30,7 +30,7 @@ class GAMRegressor(BaseProbaRegressor):
         By default a univariate spline term will be allocated for each feature.
         Can be a ``pygam`` terms expression for custom model specification.
 
-    distribution : str or pygam.Distribution, optional (default='Normal')
+    dist : str or pygam.Distribution, optional (default='Normal')
         Distribution family to use in the model.
         Supported strings (case-insensitive):
 
@@ -81,17 +81,17 @@ class GAMRegressor(BaseProbaRegressor):
     >>> y_positive = y.abs() + 1  # ensure positive targets for Poisson/Gamma
     >>>
     >>> # Normal distribution (default)
-    >>> gam_normal = GAMRegressor(distribution='Normal')
+    >>> gam_normal = GAMRegressor(dist='Normal')
     >>> gam_normal.fit(X, y)
     GAMRegressor(...)
     >>>
     >>> # Poisson distribution for count data
-    >>> gam_poisson = GAMRegressor(distribution='Poisson', link='log')
+    >>> gam_poisson = GAMRegressor(dist='Poisson', link='log')
     >>> gam_poisson.fit(X, y_positive)
     GAMRegressor(...)
     >>>
     >>> # Gamma distribution for positive continuous data
-    >>> gam_gamma = GAMRegressor(distribution='Gamma', link='log')
+    >>> gam_gamma = GAMRegressor(dist='Gamma', link='log')
     >>> gam_gamma.fit(X, y_positive)
     GAMRegressor(...)
     """
@@ -109,10 +109,16 @@ class GAMRegressor(BaseProbaRegressor):
         "tests:vm": True,
     }
 
+    # TODO (release 2.14.0)
+    # remove the 'distribution' argument from '__init__' signature
+    # remove the following 'if' check and deprecation warning
+    # de-indent the following 'else' check
+
     def __init__(
         self,
         terms="auto",
-        distribution="normal",
+        distribution="deprecated",
+        dist="normal",
         link="identity",
         max_iter=100,
         tol=1e-4,
@@ -122,6 +128,7 @@ def __init__(
     ):
         self.terms = terms
         self.distribution = distribution
+        self.dist = dist
         self.link = link
         self.max_iter = max_iter
         self.tol = tol
@@ -131,6 +138,23 @@ def __init__(
 
         super().__init__()
 
+        # handle deprecation of distribution -> dist
+        if distribution != "deprecated":
+            from warnings import warn
+
+            warn(
+                "in `GAMRegressor`, parameter 'distribution' "
+                "will be renamed to 'dist' in version 2.14.0. "
+                "To keep current behaviour and to silence this warning, "
+                "use 'dist' instead of 'distribution', "
+                "set dist explicitly via kwarg, and do not set distribution.",
+                category=DeprecationWarning,
+                stacklevel=2,
+            )
+            self._dist = distribution
+        else:
+            self._dist = dist
+
     def _fit(self, X, y):
         """Fit regressor to training data.
 
@@ -158,7 +182,7 @@ def _fit(self, X, y):
         if callbacks is None:
             callbacks = ["deviance", "diffs"]
 
-        dist_name = self._get_distribution_name(self.distribution)
+        dist_name = self._get_distribution_name(self._dist)
 
         # Map common names to skpro distribution names
         dist_map = {

skpro/regression/linear/_glm.py

@@ -22,10 +22,10 @@ class GLMRegressor(BaseProbaRegressor):
 
     Parameters
     ----------
-    family : string, default : "Normal"
-        The family parameter denotes the type of distribution
+    dist : string, default : "Normal"
+        The dist parameter denotes the type of distribution
         that will be used.
-        Available family/distributions are
+        Available distributions are
         1."Normal"
         2."Poisson"
         3."Gamma"
@@ -205,7 +205,7 @@ class GLMRegressor(BaseProbaRegressor):
         "y_inner_mtype": "pd_DataFrame_Table",
     }
 
-    def _str_to_sm_family(self, family, link):
+    def _str_to_sm_family(self, dist, link):
         """Convert the string to a statsmodel object.
 
         If the link function is also explicitly mentioned then include then
@@ -232,16 +232,22 @@ def _str_to_sm_family(self, family, link):
         if link in links:
             link_function = links[link]()
             try:
-                return sm_fmly[family](link_function)
+                return sm_fmly[dist](link_function)
             except Exception:
-                msg = "Invalid link for family, default link will be used"
+                msg = "Invalid link for distribution, default link will be used"
                 warn(msg)
 
-        return sm_fmly[family]()
+        return sm_fmly[dist]()
+
+    # TODO (release 2.14.0)
+    # remove the 'family' argument from '__init__' signature
+    # remove the following 'if' check and deprecation warning
+    # de-indent the following 'else' check
 
     def __init__(
         self,
-        family="Normal",
+        family="deprecated",
+        dist="Normal",
         link=None,
         offset_var=None,
         exposure_var=None,
@@ -262,6 +268,7 @@ def __init__(
         super().__init__()
 
         self.family = family
+        self.dist = dist
         self.link = link
         self.offset_var = offset_var
         self.exposure_var = exposure_var
@@ -279,7 +286,23 @@ def __init__(
         self.max_start_irls = max_start_irls
         self.add_constant = add_constant
 
-        self._family = self.family
+        # handle deprecation of family -> dist
+        if family != "deprecated":
+            from warnings import warn
+
+            warn(
+                "in `GLMRegressor`, parameter 'family' "
+                "will be renamed to 'dist' in version 2.14.0. "
+                "To keep current behaviour and to silence this warning, "
+                "use 'dist' instead of 'family', "
+                "set dist explicitly via kwarg, and do not set family.",
+                category=DeprecationWarning,
+                stacklevel=2,
+            )
+            self._dist = family
+        else:
+            self._dist = dist
+
         self._link = self.link
         self._offset_var = self.offset_var
         self._exposure_var = self.exposure_var
@@ -325,12 +348,12 @@ def _fit(self, X, y):
 
         # remove the offset and exposure columns which
         # was inserted to maintain the shape
-        family = self._family
+        dist = self._dist
         link = self._link
 
         # ensure numerical stability for Gamma by injecting an intercept
         self._auto_added_constant = False
-        if family == "Gamma" and not self._add_constant:
+        if dist == "Gamma" and not self._add_constant:
             self._add_constant = True
             self._auto_added_constant = True
 
@@ -341,7 +364,7 @@ def _fit(self, X, y):
 
         y_col = y.columns
 
-        sm_family = self._str_to_sm_family(family=family, link=link)
+        sm_family = self._str_to_sm_family(dist=dist, link=link)
 
         glm_estimator = GLM(
             endog=y,
@@ -434,7 +457,7 @@ def _predict(self, X):
 
         return y_pred
 
-    def _params_sm_to_skpro(self, y_predictions_df, index, columns, family):
+    def _params_sm_to_skpro(self, y_predictions_df, index, columns, dist):
         """Convert the statsmodels output to equivalent skpro distribution."""
         from skpro.distributions.gamma import Gamma
         from skpro.distributions.normal import Normal
@@ -449,8 +472,8 @@ def _params_sm_to_skpro(self, y_predictions_df, index, columns, family):
         params = {}
         skp_dist = Normal
 
-        if family in skpro_distr:
-            skp_dist = skpro_distr[family]
+        if dist in skpro_distr:
+            skp_dist = skpro_distr[dist]
 
         if skp_dist == Normal:
             y_mu = y_predictions_df["mean"].rename("mu").to_frame()
@@ -512,11 +535,11 @@ def _predict_proba(self, X):
         y_predictions_df = self.glm_fit_.get_prediction(X_).summary_frame()
 
         # convert the returned values to skpro equivalent distribution
-        family = self._family
+        dist = self._dist
         index = X_.index
         columns = y_column
 
-        y_pred = self._params_sm_to_skpro(y_predictions_df, index, columns, family)
+        y_pred = self._params_sm_to_skpro(y_predictions_df, index, columns, dist)
         return y_pred
 
     def _prep_x(self, X, offset_var, exposure_var, rtn_off_exp_arr):