Changes to Shapley tests.

tests/importance/test_shapley.py

@@ -98,9 +98,9 @@ def test_comparative_1(trainsize: int, testsize: int, method: ImportanceMethod):
 
 
 # @pytest.mark.parametrize("n_samples_train", [100, 500, 1000, 5000, 10000])
-@pytest.mark.parametrize("n_samples_train", [500, 1000, 5000, 10000])
+@pytest.mark.parametrize("n_samples_train", [1024, 2048, 4096, 8192, 16384, 32768, 65536])
 # @pytest.mark.parametrize("n_samples_test", [10, 50, 100])
-@pytest.mark.parametrize("n_samples_test", [1000])
+@pytest.mark.parametrize("n_samples_test", [1024])
 def test_neighbor_benchmark_1(n_samples_train: int, n_samples_test: int, benchmark):
     X, y = make_classification(
         n_samples=n_samples_train + n_samples_test,
@@ -122,8 +122,8 @@ def test_neighbor_benchmark_1(n_samples_train: int, n_samples_test: int, benchma
 
 
 if __name__ == "__main__":
-    n_samples_train = 10000
-    n_samples_test = 500
+    n_samples_train = 100000
+    n_samples_test = 1000
 
     X, y = make_classification(
         n_samples=n_samples_train + n_samples_test,
@@ -137,10 +137,10 @@ def test_neighbor_benchmark_1(n_samples_train: int, n_samples_test: int, benchma
     )
 
     X, X_test, y, y_test = train_test_split(X, y, train_size=n_samples_train, test_size=n_samples_test, random_state=7)
-    provenance = Provenance(units=1000)
-    provenance = provenance.fork(size=10)
+    provenance = Provenance(units=n_samples_train)
+    # provenance = provenance.fork(size=10)
 
     utility = SklearnModelAccuracy(KNeighborsClassifier(n_neighbors=1))
     importance = ShapleyImportance(method=ImportanceMethod.NEIGHBOR, utility=utility)
     importance.fit(X, y, provenance=provenance)
-    importance.score(X_test, y_test)
+    print(importance.score(X_test, y_test))