From 8ec2871e33cb51d71b7e8365f94671b509bb7186 Mon Sep 17 00:00:00 2001
From: Luc Giffon <luc.giffon@lis-lab.fr>
Date: Mon, 23 Mar 2020 11:40:46 +0100
Subject: [PATCH] optimize + fix bug kmeans forest regressor
---
.../models/kmeans_forest_regressor.py | 64 ++++++++++++++-----
1 file changed, 49 insertions(+), 15 deletions(-)
diff --git a/code/bolsonaro/models/kmeans_forest_regressor.py b/code/bolsonaro/models/kmeans_forest_regressor.py
index d0d6412..6c0e3a5 100644
--- a/code/bolsonaro/models/kmeans_forest_regressor.py
+++ b/code/bolsonaro/models/kmeans_forest_regressor.py
@@ -1,3 +1,5 @@
+import time
+
from bolsonaro.utils import tqdm_joblib
from sklearn.ensemble import RandomForestRegressor
@@ -35,29 +37,50 @@ class KMeansForestRegressor(BaseEstimator, metaclass=ABCMeta):
def fit(self, X_train, y_train, X_val, y_val):
self._estimator.fit(X_train, y_train)
- predictions = list()
- for tree in self._estimator.estimators_:
- predictions.append(tree.predict(X_train))
- predictions = np.array(predictions)
+
+ predictions_val = np.empty((len(self._estimator.estimators_), X_val.shape[0]))
+ predictions = np.empty((len(self._estimator.estimators_), X_train.shape[0]))
+ for i_tree, tree in enumerate(self._estimator.estimators_):
+ predictions_val[i_tree, :] = tree.predict(X_val)
+ predictions[i_tree, :] = tree.predict(X_train)
kmeans = KMeans(n_clusters=self._extracted_forest_size, random_state=self._models_parameters.seed).fit(predictions)
labels = np.array(kmeans.labels_)
- # For each cluster select the best tree on the validation set
- extracted_forest_sizes = list(range(self._extracted_forest_size))
- with tqdm_joblib(tqdm(total=self._extracted_forest_size, disable=True)) as prune_forest_job_pb:
- pruned_forest = Parallel(n_jobs=2)(delayed(self._prune_forest_job)(prune_forest_job_pb,
- extracted_forest_sizes[i], labels, X_val, y_val, self._score_metric)
- for i in range(self._extracted_forest_size))
-
- self._selected_trees = pruned_forest
- self._estimator.estimators_ = pruned_forest
+ # start_np_version = time.time()
+ pruned_forest_1 = list()
+ for cluster_idx in range(self._extracted_forest_size): # pourrait être parallelise
+ index_trees_cluster = np.where(labels == cluster_idx)[0]
+ predictions_val_cluster = predictions_val[index_trees_cluster] # get predictions of trees in cluster
+ if self._score_metric == mean_squared_error:
+ # compute the mean squared error of all trees at once usng numpy machinery
+ diff = predictions_val_cluster - y_val
+ squared_diff = diff ** 2
+ mean_squared_diff = np.mean(squared_diff, axis=1)
+
+ best_tree_index = np.argmin(mean_squared_diff) # get best scoring tree (the one with lowest mse)
+ pruned_forest_1.append(self._estimator.estimators_[index_trees_cluster[best_tree_index]])
+ else:
+ raise ValueError
+ # stop_np_version = time.time()
+ # print("Time np {}".format(stop_np_version - start_np_version))
+
+ # start_paralel_version = time.time()
+ # # For each cluster select the best tree on the validation set
+ # extracted_forest_sizes = list(range(self._extracted_forest_size))
+ # with tqdm_joblib(tqdm(total=self._extracted_forest_size, disable=True)) as prune_forest_job_pb:
+ # pruned_forest = Parallel(n_jobs=2)(delayed(self._prune_forest_job)(prune_forest_job_pb, extracted_forest_sizes[i], labels, X_val, y_val, self._score_metric) for i in range(self._extracted_forest_size))
+ # stop_paralel_version = time.time()
+ # print("Time paralel {}".format(stop_paralel_version - start_paralel_version))
+ # assert all([t1 is t2 for (t1, t2) in zip(pruned_forest_1, pruned_forest)])
+
+ self._selected_trees = pruned_forest_1
+ self._estimator.estimators_ = pruned_forest_1
def _prune_forest_job(self, prune_forest_job_pb, c, labels, X_val, y_val, score_metric):
index = np.where(labels == c)[0]
with tqdm_joblib(tqdm(total=len(index), disable=True)) as cluster_job_pb:
- cluster = Parallel(n_jobs=2)(delayed(self._cluster_job)(cluster_job_pb, index[i], X_val,
- y_val, score_metric) for i in range(len(index)))
+ cluster = Parallel(n_jobs=2)(delayed(self._cluster_job)(cluster_job_pb, index[i], X_val, y_val, score_metric) for i in range(len(index)))
best_tree_index = np.argmax(cluster)
prune_forest_job_pb.update()
return self._estimator.estimators_[index[best_tree_index]]
@@ -82,3 +105,14 @@ class KMeansForestRegressor(BaseEstimator, metaclass=ABCMeta):
def predict_base_estimator(self, X):
return self._estimator.predict(X)
+
+if __name__ == "__main__":
+ from sklearn import datasets
+ from bolsonaro.models.model_parameters import ModelParameters
+
+ X, y = datasets.fetch_california_housing(return_X_y=True)
+ ModelParameters(extracted_forest_size=100,
+ normalize_D=True,
+ )
+
+ k_reg = KMeansForestRegressor()
\ No newline at end of file
--
GitLab