BUILD: restore scipy 1.2 support in tests

pyproject.toml

@@ -20,7 +20,7 @@ requires = [
     "numpy          >=1.17,<2a",
     "packaging      >=20",
     "pandas         >=0.24,<2a",
-    "scipy          ~=1.4",
+    "scipy          ~=1.2",
     "shap           >=0.34,<0.41a",
     "sklearndf      ~=2.0.dev3",
 ]
@@ -76,7 +76,7 @@ numpy          = "==1.17.5"
 packaging      = "~=20.9"
 pandas         = "~=0.24.2"
 python         = ">=3.7.12,<3.8a"    # cannot use ~= due to conda bug
-scipy          = "~=1.4.1"
+scipy          = "~=1.2.3"
 shap           = "~=0.34.0"
 sklearndf      = "~=2.0.dev3"
 # additional minimum requirements of sklearndf

test/test/facet/test_selection.py

@@ -7,7 +7,7 @@
 import numpy as np
 import pandas as pd
 import pytest
-from scipy.stats import loguniform, randint
+from scipy.stats import randint, reciprocal
 from sklearn import datasets
 from sklearn.model_selection import GridSearchCV
 
@@ -160,8 +160,8 @@ def test_parameter_space(
 
     randint_3_10 = randint(3, 10)
     randint_1_32 = randint(1, 32)
-    loguniform_0_01_0_10 = loguniform(0.01, 0.1)
-    loguniform_0_05_0_10 = loguniform(0.05, 0.1)
+    reciprocal_0_01_0_10 = reciprocal(0.01, 0.1)
+    reciprocal_0_05_0_10 = reciprocal(0.05, 0.1)
 
     # parameter space 1
 
@@ -171,9 +171,9 @@ def test_parameter_space(
     )
     ps_1_name = "rf_regressor"
     ps_1 = ParameterSpace(pipeline_1, name=ps_1_name)
-    ps_1.regressor.min_weight_fraction_leaf = loguniform_0_01_0_10
+    ps_1.regressor.min_weight_fraction_leaf = reciprocal_0_01_0_10
     ps_1.regressor.max_depth = randint_3_10
-    ps_1.regressor.min_samples_leaf = loguniform_0_05_0_10
+    ps_1.regressor.min_samples_leaf = reciprocal_0_05_0_10
 
     with pytest.raises(
         AttributeError,
@@ -239,9 +239,9 @@ def regressor_repr(model: Id):
             Id.ParameterSpace(
                 regressor_repr(Id.RandomForestRegressorDF),
                 **{
-                    "regressor.min_weight_fraction_leaf": (Id.loguniform(0.01, 0.1)),
+                    "regressor.min_weight_fraction_leaf": (Id.reciprocal(0.01, 0.1)),
                     "regressor.max_depth": Id.randint(3, 10),
-                    "regressor.min_samples_leaf": (Id.loguniform(0.05, 0.1)),
+                    "regressor.min_samples_leaf": (Id.reciprocal(0.05, 0.1)),
                 },
             ),
             Id.ParameterSpace(
@@ -261,8 +261,8 @@ def regressor_repr(model: Id):
             "candidate": [pipeline_1],
             "candidate_name": [ps_1_name],
             "candidate__regressor__max_depth": randint_3_10,
-            "candidate__regressor__min_samples_leaf": loguniform_0_05_0_10,
-            "candidate__regressor__min_weight_fraction_leaf": loguniform_0_01_0_10,
+            "candidate__regressor__min_samples_leaf": reciprocal_0_05_0_10,
+            "candidate__regressor__min_weight_fraction_leaf": reciprocal_0_01_0_10,
         },
         {
             "candidate": [pipeline_2],
@@ -277,10 +277,10 @@ def regressor_repr(model: Id):
             "my_prefix__candidate": [pipeline_1],
             "my_prefix__candidate_name": [ps_1_name],
             "my_prefix__candidate__regressor__max_depth": randint_3_10,
-            "my_prefix__candidate__regressor__min_samples_leaf": loguniform_0_05_0_10,
+            "my_prefix__candidate__regressor__min_samples_leaf": reciprocal_0_05_0_10,
             (
                 "my_prefix__candidate__regressor__min_weight_fraction_leaf"
-            ): loguniform_0_01_0_10,
+            ): reciprocal_0_01_0_10,
         },
         {
             "my_prefix__candidate": [pipeline_2],