WIP: Resolve "coherence des arbres de predictions"

Closes #20 (closed)

code/bolsonaro/trainer.py

@@ -77,7 +77,7 @@ class Trainer(object):
         else:
             raise ValueError("Unknown specified subsets_used parameter '{}'".format(model.models_parameters.subsets_used))
 
-    def train(self, model, extracted_forest_size=None):
+    def train(self, model, extracted_forest_size=None, seed=None):
         """
         :param model: An instance of either RandomForestRegressor, RandomForestClassifier, OmpForestRegressor,
             OmpForestBinaryClassifier, OmpForestMulticlassClassifier.
@@ -88,6 +88,7 @@ class Trainer(object):
         if type(model) in [RandomForestRegressor, RandomForestClassifier]:
             if extracted_forest_size is not None:
                 estimators_index = np.arange(len(model.estimators_))
+                np.random.seed(seed)
                 np.random.shuffle(estimators_index)
                 choosen_estimators = estimators_index[:extracted_forest_size]
                 model.estimators_ = np.array(model.estimators_)[choosen_estimators]
@@ -141,6 +142,30 @@ class Trainer(object):
             result = self._base_regression_score_metric(y_true, y_pred)
         return result
 
+    def compute_preds_coherence(self, model, X):
+        from sklearn.preprocessing import normalize
+        import itertools
+
+        if type(model) in [OmpForestRegressor, SimilarityForestRegressor, KMeansForestRegressor, EnsembleSelectionForestRegressor,
+            OmpForestBinaryClassifier, OmpForestMulticlassClassifier]:
+            estimators = model.forest
+            estimators = np.asarray(estimators)[model._omp.coef_ != 0]
+        elif type(model) in [RandomForestRegressor, RandomForestClassifier]:
+            estimators = model.estimators_
+
+        predictions = np.array([tree.predict(X) for tree in estimators])
+
+        predictions = normalize(predictions)
+
+        """similarities = list()
+        for ti, tj in list(itertools.combinations(predictions, 2)):
+            similarities.append(np.abs(ti @ tj))
+        coherence = np.max(np.asarray(similarities))"""
+
+        coherence = np.max(np.abs((predictions @ predictions.T - np.eye(len(predictions)))))
+
+        return coherence
+
     def compute_results(self, model, models_dir):
         """
         :param model: Object with
@@ -173,7 +198,8 @@ class Trainer(object):
             dev_score_base=self.__score_func_base(model, self._dataset.X_dev, self._dataset.y_dev),
             test_score_base=self.__score_func_base(model, self._dataset.X_test, self._dataset.y_test),
             score_metric=self._score_metric_name,
-            base_score_metric=self._base_score_metric_name
+            base_score_metric=self._base_score_metric_name,
+            coherence=self.compute_preds_coherence(model, self._dataset.X_train)
         )
         results.save(models_dir)
         self._logger.info("Base performance on test: {}".format(results.test_score_base))