diff --git a/code/bolsonaro/models/ensemble_selection_forest_regressor.py b/code/bolsonaro/models/ensemble_selection_forest_regressor.py
index 1edbab0479b4553e290eb6eafcb1f1ad84a984e9..fb399d19e06efe61a84ec9956800b74a1709d8bb 100644
--- a/code/bolsonaro/models/ensemble_selection_forest_regressor.py
+++ b/code/bolsonaro/models/ensemble_selection_forest_regressor.py
@@ -30,6 +30,8 @@ class EnsembleSelectionForest(ForestPruningSOTA, metaclass=ABCMeta):
         mean_so_far = val_predictions[idx_best_score]
         while nb_selected_trees < self._extracted_forest_size:
             # every new tree is selected with replacement as specified in the base paper
+
+            # this matrix contains  at each line the predictions of the previous subset + the corresponding tree of the line
             # mean update formula: u_{t+1} = (n_t * u_t + x_t) / (n_t + 1)
             mean_prediction_subset_with_extra_tree = (nb_selected_trees * mean_so_far + val_predictions) / (nb_selected_trees + 1)
             predictions_subset_with_extra_tree = self._activation(mean_prediction_subset_with_extra_tree)
@@ -37,6 +39,7 @@ class EnsembleSelectionForest(ForestPruningSOTA, metaclass=ABCMeta):
             idx_best_extra_tree = self._best_score_idx(scores_subset_with_extra_tree)
             lst_pruned_forest.append(self._base_estimator.estimators_[idx_best_extra_tree])
 
+            # update new mean prediction
             mean_so_far = mean_prediction_subset_with_extra_tree[idx_best_extra_tree]
             nb_selected_trees += 1