Merge pull request #33 from fidelity/version1.11.2

version 1.11.2 update

CHANGELOG.txt

@@ -2,6 +2,13 @@
 MABWiser CHANGELOG
 =====================
 
+-------------------------------------------------------------------------------
+April, 15, 2021 1.11.2
+-------------------------------------------------------------------------------
+minor:
+- Updated dtype from np.int to int in base_mab.py line 294 to resolve deprecation warning from numpy>=1.20.0
+- Updated sklearn version dependency to >=0.24.0 because of changes in the KMeans implementation
+
 -------------------------------------------------------------------------------
 September, 14, 2020 1.11.1
 -------------------------------------------------------------------------------

README.md

@@ -56,7 +56,6 @@ MABWiser is available to install as `pip install mabwiser`. It can also be insta
 
 Please submit bug reports and feature requests as [Issues](https://github.com/fidelity/mabwiser/issues).
 
-
 ## Citation
 
 If you use MABWiser in a publication, please cite it as:

mabwiser/_version.py

@@ -3,5 +3,5 @@
 
 __author__ = "FMR LLC"
 __email__ = "[email protected]"
-__version__ = "1.11.1"
+__version__ = "1.11.2"
 __copyright__ = "Copyright (C), FMR LLC"

mabwiser/base_mab.py

@@ -191,7 +191,7 @@ def _partition_contexts(self, n_contexts: int):
         n_jobs = self._effective_jobs(n_contexts, self.n_jobs)
 
         # Partition contexts between jobs
-        n_contexts_per_job = np.full(n_jobs, n_contexts // n_jobs, dtype=np.int)
+        n_contexts_per_job = np.full(n_jobs, n_contexts // n_jobs, dtype=int)
         n_contexts_per_job[:n_contexts % n_jobs] += 1
         starts = np.cumsum(n_contexts_per_job)
 

requirements.txt

@@ -1,6 +1,6 @@
 joblib
 numpy
 pandas
-scikit-learn>=0.22.0
+scikit-learn>=0.24.0
 scipy
-seaborn>=0.9.0
+seaborn>=0.9.0

tests/test_clusters.py

@@ -102,7 +102,7 @@ def test_greedy0_n3(self):
                                  num_run=1,
                                  is_predict=True)
 
-        self.assertListEqual(arms, [2, 2])
+        self.assertListEqual(arms, [2, 3])
         self.assertEqual(mab._imp.kmeans.n_clusters, 3)
 
     def test_greedy1_n3(self):
@@ -231,7 +231,7 @@ def binarize(arm, reward):
                                  is_predict=True)
 
         self.assertTrue(mab._imp.lp_list[0].is_contextual_binarized)
-        self.assertListEqual(arms, [3, 4])
+        self.assertListEqual(arms, [3, 3])
         self.assertEqual(len(mab._imp.decisions), 10)
         self.assertEqual(len(mab._imp.rewards), 10)
         self.assertEqual(len(mab._imp.contexts), 10)
@@ -271,7 +271,7 @@ def binarize(arm, reward):
                                  is_predict=True)
 
         self.assertTrue(mab._imp.lp_list[0].is_contextual_binarized)
-        self.assertListEqual(arms, [3, 4])
+        self.assertListEqual(arms, [3, 3])
         self.assertEqual(len(mab._imp.decisions), 10)
         self.assertEqual(len(mab._imp.rewards), 10)
         self.assertEqual(len(mab._imp.contexts), 10)

tests/test_parallel.py

@@ -236,7 +236,7 @@ def test_UCB1_c2(self):
                                 is_predict=True,
                                 n_jobs=1)
 
-        self.assertEqual(arm,  [[3, 3, 3, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
+        self.assertEqual(arm,  [[3, 3, 1, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
 
         arm, mab = self.predict(arms=[1, 2, 3, 4],
                                 decisions=[1, 1, 1, 2, 2, 2, 3, 3, 3, 4],
@@ -250,7 +250,7 @@ def test_UCB1_c2(self):
                                 is_predict=True,
                                 n_jobs=2)
 
-        self.assertEqual(arm,  [[3, 3, 3, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
+        self.assertEqual(arm,  [[3, 3, 1, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
 
         arm, mab = self.predict(arms=[1, 2, 3, 4],
                                 decisions=[1, 1, 1, 2, 2, 2, 3, 3, 3, 4],
@@ -264,7 +264,7 @@ def test_UCB1_c2(self):
                                 is_predict=True,
                                 n_jobs=100)
 
-        self.assertEqual(arm,  [[3, 3, 3, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
+        self.assertEqual(arm,  [[3, 3, 1, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
 
     def test_greedy1_k2(self):
         rng = np.random.RandomState(seed=7)
@@ -576,7 +576,7 @@ def test_thompson_n3(self):
                                  is_predict=True,
                                  n_jobs=-1)
 
-        self.assertListEqual(arms, [1, 1, 3, 1, 1, 1, 1, 2, 1, 2])
+        self.assertListEqual(arms, [2, 1, 3, 3, 3, 2, 2, 3, 2, 3])
 
     def test_thompson_a2(self):
 
@@ -879,7 +879,7 @@ def test_UCB1_c2_backend(self):
                                 n_jobs=2,
                                 backend=None)
 
-        self.assertEqual(arm, [[3, 3, 3, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
+        self.assertEqual(arm, [[3, 3, 1, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
 
         arm, mab = self.predict(arms=[1, 2, 3, 4],
                                 decisions=[1, 1, 1, 2, 2, 2, 3, 3, 3, 4],
@@ -894,7 +894,7 @@ def test_UCB1_c2_backend(self):
                                 n_jobs=2,
                                 backend='loky')
 
-        self.assertEqual(arm, [[3, 3, 3, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
+        self.assertEqual(arm, [[3, 3, 1, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
 
         arm, mab = self.predict(arms=[1, 2, 3, 4],
                                 decisions=[1, 1, 1, 2, 2, 2, 3, 3, 3, 4],
@@ -909,7 +909,7 @@ def test_UCB1_c2_backend(self):
                                 n_jobs=2,
                                 backend='threading')
 
-        self.assertEqual(arm, [[3, 3, 3, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
+        self.assertEqual(arm, [[3, 3, 1, 1, 1, 1, 3, 1, 3, 3] for _ in range(5)])
 
     def test_greedy1_k2_backend(self):
         rng = np.random.RandomState(seed=7)