...

pyxai/examples/DT/builder-rectify3.py

@@ -18,9 +18,7 @@
     "binary": ["PP", "R"],
 }
 
-print("bob = (25, 1, 1):")
-bob = (25, 1, 1)
-explainer = Explainer.initialize(tree, instance=bob, features_type=loan_types)
+explainer = Explainer.initialize(tree, instance=(25, 1, 1), features_type=loan_types)
 
 print("binary representation: ", explainer.binary_representation)
 print("target_prediction:", explainer.target_prediction)
@@ -29,6 +27,6 @@
 #For him/her, the following classification rule must be obeyed:
 #whenever the annual income of the client is lower than 30,
 #the demand should be rejected
-rectified_model = explainer.rectify(decision_rule=(-1, ), label=0) 
+rectified_model = explainer.rectify(conditions=(-1, ), label=0) 
 
 assert (0, (1, 0, (4, (3, 0, 1), 1))) == rectified_model.raw_data_for_CPP(), "The rectified model is not good."

pyxai/sources/core/explainer/explainerDT.py

@@ -316,10 +316,10 @@ def simplify_theory(self, tree_rectified):
             return tree_rectified
         return tree_rectified
 
-    def rectify(self, *, decision_rule, label):
+    def rectify(self, *, conditions, label):
         """
-        Rectify the Decision Tree (self._tree) of the explainer according to a `decision_rule` and a `label`.
-        Simplify the final tree (the theory can help to eliminate some nodes).
+        Rectify the Decision Tree (self._tree) of the explainer according to a `conditions` and a `label`.
+        Simplify the model (the theory can help to eliminate some nodes).
 
         Args:
             decision_rule (list or tuple): A decision rule in the form of list of literals (binary variables representing the conditions of the tree). 
@@ -329,7 +329,7 @@ def rectify(self, *, decision_rule, label):
         """
         Tools.verbose("")
         Tools.verbose("-------------- Rectify information:")
-        tree_decision_rule = self._tree.decision_rule_to_tree(decision_rule)
+        tree_decision_rule = self._tree.decision_rule_to_tree(conditions)
         Tools.verbose("Desision Rule Number of nodes:", tree_decision_rule.n_nodes())
         Tools.verbose("Model Number of nodes:", self._tree.n_nodes())
         if label == 1:

pyxai/sources/solvers/MAXSAT/OPENWBOSolver.py

@@ -33,7 +33,7 @@ def solve(self, *, time_limit=0):
         output_str = [line.split(" ") for line in p.stdout.split(os.linesep) if len(line) > 0 and line[0] == "v"]
         if len(output_str) == 0:
             return p.stderr, None, time_used
-
+        
         status = [line.split(" ")[1] for line in p.stdout.split(os.linesep) if len(line) > 0 and line[0] == "s"][0]
         model = [int(lit) for lit in output_str[0] if lit != 'v' and lit != '']
         return status, model, Explainer.TIMEOUT if status == "SATISFIABLE" else time_used

pyxai/tests/functionality/Rectify.py

@@ -2,7 +2,7 @@
 import math
 
 
-Tools.set_verbose(0)
+#Tools.set_verbose(0)
 
 import unittest
 class TestRectify(unittest.TestCase):
@@ -20,8 +20,8 @@ def test_rectify_1(self):
         explainer = Explainer.initialize(model, features_type=loan_types)
 
         #Alice’s expertise can be represented by the formula T = ((x1 ∧ not x3) ⇒ y) ∧ (not x2 ⇒ not y) encoding her two decision rules
-        explainer.rectify(decision_rule=(1, -3), label=1)  #(x1 ∧ not x3) ⇒ y
-        explainer.rectify(decision_rule=(-2, ), label=0)  #not x2 ⇒ not y
+        explainer.rectify(conditions=(1, -3), label=1)  #(x1 ∧ not x3) ⇒ y
+        explainer.rectify(conditions=(-2, ), label=0)  #not x2 ⇒ not y
 
         rectified_model = explainer.get_model().raw_data_for_CPP()
 
@@ -57,7 +57,7 @@ def test_rectify_2(self):
         minimal = explainer.minimal_sufficient_reason()
 
 
-        explainer.rectify(decision_rule=minimal, label=1) 
+        explainer.rectify(conditions=minimal, label=1) 
         rectified_model = explainer.get_model().raw_data_for_CPP()
         self.assertEqual(rectified_model, (0, (1, (2, (5, (6, 1, 0), 1), 1), 1)))
 
@@ -84,7 +84,7 @@ def test_rectify_4(self):
         #For him/her, the following classification rule must be obeyed:
         #whenever the annual income of the client is lower than 30,
         #the demand should be rejected
-        rectified_model = explainer.rectify(decision_rule=(-1, ), label=0) 
+        rectified_model = explainer.rectify(conditions=(-1, ), label=0) 
 
         self.assertEqual(rectified_model.raw_data_for_CPP(), (0, (1, 0, (4, (3, 0, 1), 1))))
 
@@ -104,16 +104,17 @@ def test_rectify_3(self):
 
         compas_types = {
             "numerical": ["Number_of_Priors"],
-            "binary": ["Misdemeanor", "score_factor", "Age_Above_FourtyFive", "Age_Below_TwentyFive", "Female"],
-            "categorical": {"Origin*": ["African_American", "Asian", "Hispanic", "Native_American", "Other"]}
+            "binary": ["Misdemeanor", "score_factor", "Female"],
+            "categorical": {"Origin*": ["African_American", "Asian", "Hispanic", "Native_American", "Other"],
+                            "Age*": ["Above_FourtyFive", "Below_TwentyFive"]}
         }
 
 
         explainer = Explainer.initialize(model, instance=instance, features_type=compas_types)
-        minimal_reason = explainer.sufficient_reason(n=1)
-        #print("explanation:", minimal_reason)
-        #print("explanation:", explainer.to_features(minimal_reason))
-        model = explainer.rectify(decision_rule=minimal_reason, label=1)
+        minimal_reason = explainer.minimal_sufficient_reason(n=1)
+        print("explanation:", minimal_reason)
+        print("explanation:", explainer.to_features(minimal_reason))
+        model = explainer.rectify(conditions=minimal_reason, label=1)
 
         self.assertEqual(model.predict_instance(instance), 1)