resolving bug of get_selected_index

Author: bishwamittra

documentation.ipynb

@@ -133,7 +133,7 @@
      "output_type": "stream",
      "name": "stdout",
      "text": [
-      "\nTraining started for batch:  1\n- number of soft clauses:  93\n- number of Boolean variables: 157\n- number of hard and soft clauses: 863\n\n\nBatch tarining complete\n- number of literals in the rule: 2\n- number of training errors:    3 out of 49\n\nTraining started for batch:  2\n- number of soft clauses:  95\n- number of Boolean variables: 161\n- number of hard and soft clauses: 890\n\n\nBatch tarining complete\n- number of literals in the rule: 4\n- number of training errors:    1 out of 51\n"
+      "(100, 11)\n\nTraining started for batch:  1\n- number of soft clauses:  93\n- number of Boolean variables: 157\n- number of hard and soft clauses: 863\n\n\nBatch tarining complete\n- number of literals in the rule: 2\n- number of training errors:    3 out of 49\n\nTraining started for batch:  2\n- number of soft clauses:  95\n- number of Boolean variables: 161\n- number of hard and soft clauses: 890\n\n\nBatch tarining complete\n- number of literals in the rule: 4\n- number of training errors:    1 out of 51\n(100, 11)\n"
      ]
     }
    ],
@@ -150,7 +150,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 6,
+   "execution_count": 8,
    "metadata": {
     "tags": []
    },
@@ -168,8 +168,7 @@
     "print(classification_report(y_train, model.predict(X_train), target_names=['0','1']))\n",
     "print()\n",
     "print(\"test report: \")\n",
-    "print(classification_report(y_test, model.predict(X_test), target_names=['0','1']))\n",
-    "\n"
+    "print(classification_report(y_test, model.predict(X_test), target_names=['0','1']))\n"
    ]
   },
   {
@@ -181,7 +180,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 7,
+   "execution_count": 9,
    "metadata": {
     "tags": []
    },
@@ -212,7 +211,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 8,
+   "execution_count": 10,
    "metadata": {
     "tags": []
    },
@@ -223,7 +222,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 9,
+   "execution_count": 13,
    "metadata": {
     "tags": []
    },
@@ -232,7 +231,7 @@
      "output_type": "stream",
      "name": "stdout",
      "text": [
-      "training report: \n              precision    recall  f1-score   support\n\n           0       0.93      0.98      0.96        65\n           1       0.97      0.86      0.91        35\n\n    accuracy                           0.94       100\n   macro avg       0.95      0.92      0.93       100\nweighted avg       0.94      0.94      0.94       100\n\n\ntest report: \n              precision    recall  f1-score   support\n\n           0       0.97      1.00      0.99        35\n           1       1.00      0.93      0.97        15\n\n    accuracy                           0.98        50\n   macro avg       0.99      0.97      0.98        50\nweighted avg       0.98      0.98      0.98        50\n\n\n( not sepal length >=  7.05 AND petal width = (0.8 - 1.75) ) OR \n( petal length = (2.45 - 4.75))\n\n[[-2, 9], [6]]\n"
+      "training report: \n              precision    recall  f1-score   support\n\n           0       0.93      0.98      0.96        65\n           1       0.97      0.86      0.91        35\n\n    accuracy                           0.94       100\n   macro avg       0.95      0.92      0.93       100\nweighted avg       0.94      0.94      0.94       100\n\n\ntest report: \n              precision    recall  f1-score   support\n\n           0       0.97      1.00      0.99        35\n           1       1.00      0.93      0.97        15\n\n    accuracy                           0.98        50\n   macro avg       0.99      0.97      0.98        50\nweighted avg       0.98      0.98      0.98        50\n\n\nRule:->\n( not sepal length >=  7.05 AND petal width = (0.8 - 1.75) ) OR \n( petal length = (2.45 - 4.75))\n\nOriginal features:\n['sepal length <  5.45', 'sepal length = (5.45 - 7.05)', 'sepal length >=  7.05', 'sepal width <  2.95', 'sepal width >=  2.95', 'petal length <  2.45', 'petal length = (2.45 - 4.75)', 'petal length >=  4.75', 'petal width <  0.8', 'petal width = (0.8 - 1.75)', 'petal width >=  1.75']\n\nIn the learned rule, show original index in the feature list with phase (1: original, -1: complemented)\n[[(2, -1), (9, 1)], [(6, 1)]]\n"
      ]
     }
    ],
@@ -246,9 +245,11 @@
     "print(\"test report: \")\n",
     "print(classification_report(y_test, model.predict(X_test), target_names=['0','1']))\n",
     "\n",
-    "print()\n",
+    "print(\"\\nRule:->\")\n",
     "print(model.get_rule(features))\n",
-    "print()\n",
+    "print(\"\\nOriginal features:\")\n",
+    "print(features)\n",
+    "print(\"\\nIn the learned rule, show original index in the feature list with phase (1: original, -1: complemented)\")\n",
     "print(model.get_selected_column_index())"
    ]
   },
@@ -266,7 +267,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 10,
+   "execution_count": 14,
    "metadata": {},
    "outputs": [],
    "source": [
@@ -275,7 +276,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 11,
+   "execution_count": 15,
    "metadata": {
     "tags": []
    },
@@ -284,7 +285,7 @@
      "output_type": "stream",
      "name": "stdout",
      "text": [
-      "training report: \n              precision    recall  f1-score   support\n\n           0       0.96      0.98      0.97        65\n           1       0.97      0.91      0.94        35\n\n    accuracy                           0.96       100\n   macro avg       0.96      0.95      0.96       100\nweighted avg       0.96      0.96      0.96       100\n\n\ntest report: \n              precision    recall  f1-score   support\n\n           0       0.97      1.00      0.99        35\n           1       1.00      0.93      0.97        15\n\n    accuracy                           0.98        50\n   macro avg       0.99      0.97      0.98        50\nweighted avg       0.98      0.98      0.98        50\n\n"
+      "training report: \n              precision    recall  f1-score   support\n\n           0       0.91      0.31      0.46        65\n           1       0.42      0.94      0.58        35\n\n    accuracy                           0.53       100\n   macro avg       0.67      0.63      0.52       100\nweighted avg       0.74      0.53      0.50       100\n\n\ntest report: \n              precision    recall  f1-score   support\n\n           0       0.90      0.26      0.40        35\n           1       0.35      0.93      0.51        15\n\n    accuracy                           0.46        50\n   macro avg       0.62      0.60      0.45        50\nweighted avg       0.73      0.46      0.43        50\n\n"
      ]
     }
    ],
@@ -310,7 +311,7 @@
   },
   {
    "cell_type": "code",
-   "execution_count": 12,
+   "execution_count": 16,
    "metadata": {
     "tags": []
    },
@@ -319,7 +320,7 @@
      "output_type": "stream",
      "name": "stdout",
      "text": [
-      "Learned rule is: \n\nAn Iris flower is predicted as Iris Versicolor if\n[ (  sepal width >=  2.95  + petal length = (2.45 - 4.75)  + petal width = (0.8 - 1.75)  + not sepal length >=  7.05   )>= 3  ] +\n[ ( )>= 0  ]  >= 2\n\nThrehosld on clause: 2\nThreshold on literals: (this is a list where the entries denote threholds on literals on all clauses)\n[3, 0]\n"
+      "Learned rule is: \n\nAn Iris flower is predicted as Iris Versicolor if\n[ (  petal width = (0.8 - 1.75)  + not sepal length >=  7.05   )>= 1  ] +\n[ (  sepal width >=  2.95  + petal length = (2.45 - 4.75)   )>= 1  ]  >= 2\n\nThrehosld on clause: 2\nThreshold on literals: (this is a list where the entries denote threholds on literals on all clauses)\n[1, 1]\n"
      ]
     }
    ],
@@ -329,7 +330,7 @@
     "print(\"An Iris flower is predicted as Iris Versicolor if\")\n",
     "print(rule)\n",
     "print(\"\\nThrehosld on clause:\", model.get_threshold_clause())\n",
-    "print(\"Threshold on literals: (this is a list where the entries denote threholds on literals on all clauses)\")\n",
+    "print(\"Threshold on literals: (this is a list where entries denote threholds on literals on all clauses)\")\n",
     "print(model.get_threshold_literal())"
    ]
   },

pyrulelearn/imli.py

@@ -134,12 +134,14 @@ def get_selected_column_index(self):
         for index_list in temp:
             each_level_index = []
             for index in index_list:
+                phase = 1
                 actual_feature_len = int(self.numFeatures/2)
                 if(self.ruleType == "DNF"):
                     index = index - actual_feature_len if index >= actual_feature_len else index + actual_feature_len
                 if(index >= actual_feature_len):
-                    index = -1 * (index - actual_feature_len)
-                each_level_index.append(index)
+                    index = index - actual_feature_len
+                    phase = -1
+                each_level_index.append((index, phase))
             result.append(each_level_index)
 
         return result
@@ -775,6 +777,7 @@ def __generateWcnfFile(self, AMatrix, yVector, xSize, WCNFFile,
     def get_rule(self, features):
 
         if(2 * len(features) == self.numFeatures):
+            features = [str(feature) for feature in features]
             features += ["not " + str(feature) for feature in features]
 
         assert len(features) == self.numFeatures

setup.py

@@ -10,15 +10,15 @@
 setup(
   name = 'pyrulelearn',
   packages = ['pyrulelearn'],
-  version = 'v1.0.8',
+  version = 'v1.0.9',
   license='MIT',
   description = 'This library can be used to generate interpretable classification rules expressed as CNF/DNF and relaxed-CNF',
   long_description=long_description,
   long_description_content_type='text/markdown',
   author = 'Bishwamittra Ghosh',
   author_email = 'bishwamittra.ghosh@gmail.com',
   url = 'https://github.com/meelgroup/MLIC',
-  download_url = 'https://github.com/meelgroup/MLIC/archive/v1.0.7.tar.gz',
+  download_url = 'https://github.com/meelgroup/MLIC/archive/v1.0.8.tar.gz',
   keywords = ['Classification Rules', 'Interpretable Rules', 'CNF Classification Rules', 'DNF Classification Rules','MaxSAT-based Rule Learning'],   # Keywords that define your package best
   classifiers=[
     'Development Status :: 3 - Alpha',