MABWiser update (#24)

.github/workflows/ci.yml

@@ -10,10 +10,11 @@ on:
 
 jobs:
   Test:
-    runs-on: ubuntu-latest
+    runs-on: ${{ matrix.os }}
     strategy:
       matrix:
-        python-version: [3.7, 3.8, 3.9]
+        python-version: ['3.7', '3.8', '3.9', '3.10']
+        os: [ubuntu-latest, macos-latest, windows-latest]
     steps:
       - uses: actions/checkout@v2
       - name: Set up Python ${{ matrix.python-version }}

CHANGELOG.txt

@@ -2,6 +2,15 @@
 CHANGELOG
 =========
 
+-------------------------------------------------------------------------------
+February, 23, 2023 1.1.0
+-------------------------------------------------------------------------------
+
+Major:
+- Updated requirements to use mabwiser>=2.7 to benefit from enhancements,
+  including vectorized predict for Linear policies and tracking of arm status.
+- Fixed tests due to changes in random seeding for Linear policies.
+
 -------------------------------------------------------------------------------
 August, 16, 2022 1.0.3
 -------------------------------------------------------------------------------

requirements.txt

@@ -4,6 +4,6 @@ scikit-learn>=0.24.0
 scipy
 matplotlib
 seaborn>=0.1.0
-mabwiser>=2.4.0
+mabwiser>=2.7.0
 jurity>=1.3.2
 spock-config>=2.1.4

tests/test_rec.py

@@ -611,13 +611,13 @@ def test_recommend_lin_greedy(self):
                                                      [0, 2, 1, 0, 0]],
                                     top_k=2,
                                     seed=123456)
-        self.assertEqual(results[0], [[3, 1], [1, 3]])
+        self.assertEqual(results[0], [[3, 1], [2, 3]])
         self.assertListAlmostEqual(results[1][0], [0.6504125435586658, 0.5240639631785098])
-        self.assertListAlmostEqual(results[1][1], [0.7114885562660387, 0.5354020923417662])
+        self.assertListAlmostEqual(results[1][1], [0.7221703184615302, 0.7121913829791641])
 
         # No scores
         results = rec.recommend([[0, 1, 2, 3, 5], [1, 1, 1, 1, 1]], return_scores=False)
-        self.assertEqual(results, [[3, 1], [2, 3]])
+        self.assertEqual(results, [[2, 3], [2, 1]])
 
     def test_recommend_lin_ucb(self):
         results, rec = self.predict(arms=[1, 2, 3],
@@ -653,13 +653,13 @@ def test_recommend_lin_ts(self):
                                                      [0, 2, 1, 0, 0]],
                                     top_k=2,
                                     seed=123456)
-        self.assertEqual(results[0], [[2, 3], [3, 1]])
-        self.assertListAlmostEqual(results[1][0], [0.9571299309237765, 0.7351400505873965])
-        self.assertListAlmostEqual(results[1][1], [0.8548770839301622, 0.7726819822665895])
+        self.assertEqual(results[0], [[3, 2], [1, 3]])
+        self.assertListAlmostEqual(results[1][0], [0.6393327956234724, 0.6113795857188596])
+        self.assertListAlmostEqual(results[1][1], [0.9231640190086698, 0.9093145340785204])
 
         # No scores
         results = rec.recommend([[0, 1, 2, 3, 5], [1, 1, 1, 1, 1]], return_scores=False)
-        self.assertEqual(results, [[2, 3], [1, 3]])
+        self.assertEqual(results, [[3, 1], [1, 3]])
 
     def test_recommend_clusters_ts(self):
         results, rec = self.predict(arms=[1, 2, 3],