remove old similarity parameters

Author: Florian Gardin

skrules/skope_rules.py

@@ -14,7 +14,6 @@
 
 INTEGER_TYPES = (numbers.Integral, np.integer)
 
-
 class SkopeRules(BaseEstimator):
     """ An easy-interpretable classifier optimizing simple logical rules.
 
@@ -35,11 +34,6 @@ class SkopeRules(BaseEstimator):
         The number of base estimators (rules) to use for prediction. More are
         built before selection. All are available in the estimators_ attribute.
 
-    similarity_thres : float, optional (default=0.99)
-        Similarity threshold between rules. Rules too similar
-        (> similarity_thres) are fused. The similarity between two rules is
-        computed according to the formula `# {intersection} / # {union}`.
-
     max_samples : int or float, optional (default=.8)
         The number of samples to draw from X to train each decision tree, from
         which rules are generated and selected.
@@ -66,7 +60,8 @@ class SkopeRules(BaseEstimator):
         than min_samples_split samples.
 
     max_depth_duplication : integer or None, optional (default=3)
-        The maximum depth of the decision tree for rule deduplication, if None then no deduplication occurs.
+        The maximum depth of the decision tree for rule deduplication,
+        if None then no deduplication occurs.
 
     max_features : int, float, string or None, optional (default="auto")
         The number of features considered (by each decision tree) when looking
@@ -140,13 +135,12 @@ def __init__(self,
                  precision_min=0.5,
                  recall_min=0.01,
                  n_estimators=10,
-                 similarity_thres=0.95,
                  max_samples=.8,
                  max_samples_features=1.,
                  bootstrap=False,
                  bootstrap_features=False,
                  max_depth=3,
-                 max_depth_duplication=3,
+                 max_depth_duplication=None,
                  max_features=1.,
                  min_samples_split=2,
                  n_jobs=1,
@@ -156,7 +150,6 @@ def __init__(self,
         self.recall_min = recall_min
         self.feature_names = feature_names
         self.n_estimators = n_estimators
-        self.similarity_thres = similarity_thres
         self.max_samples = max_samples
         self.max_samples_features = max_samples_features
         self.bootstrap = bootstrap
@@ -214,11 +207,6 @@ def fit(self, X, y, sample_weight=None):
                  % set(self.classes_))
             y = (y > 0)
 
-        # ensure similarity_thres is in (0., 1.]:
-        if not (0. < self.similarity_thres <= 1.):
-            raise ValueError("similarity_thres must be in (0, 1], got %r"
-                             % self.similarity_thres)
-
         # ensure that max_samples is in [1, n_samples]:
         n_samples = X.shape[0]
 
@@ -640,8 +628,10 @@ def split_with_best_feature(rules, depth, exceptions=[]):
                     rules_splitted[2].append(rule)
 
             # Choose best term
-            return [split_with_best_feature(ruleset, depth-1, exceptions=exceptions+[most_represented_term]) for ruleset in rules_splitted]
-
+            return [split_with_best_feature(ruleset,
+                                            depth-1,
+                                            exceptions=exceptions+[most_represented_term])
+                    for ruleset in rules_splitted]
 
         def breadth_first_search(rules, leaves=None):
             if len(rules) == 0 or not isinstance(rules[0], list):

skrules/tests/test_skope_rules.py

@@ -79,12 +79,6 @@ def test_skope_rules_error():
     y = iris.target
     y = (y != 0)
 
-    # Test similarity_thres:
-    assert_raises(ValueError,
-                  SkopeRules(similarity_thres=2).fit, X, y)
-    assert_raises(ValueError,
-                  SkopeRules(similarity_thres=0).fit, X, y)
-
     # Test max_samples
     assert_raises(ValueError,
                   SkopeRules(max_samples=-1).fit, X, y)
@@ -176,7 +170,7 @@ def test_performances():
 
 
 def test_similarity_tree():
-    ## Test that rules are well splitted
+    # Test that rules are well splitted
     rules = [("a <= 2 and b > 45 and c <= 3 and a > 4", (1, 1, 0)),
              ("a <= 2 and b > 45 and c <= 3 and a > 4", (1, 1, 0)),
              ("a > 2 and b > 45", (0.5, 0.3, 0)),