def test_classification_statistics_encodable(self):
c1 = [{"1":1, "2":2 }, "A" ]
c2 = [{"1":3, "2":4 }, "B" ]
env = SupervisedSimulation(*zip(*[c1,c2]*10))
row = ClassEnvironmentTask().process(env,env.read())
json.dumps(row)
def test_classification_statistics_encodable_sans_sklearn(self):
with unittest.mock.patch('importlib.import_module', side_effect=ImportError()):
c1 = [{"1":1, "2":2 }, "A" ]
c2 = [{"1":3, "2":4 }, "B" ]
simulation = SupervisedSimulation(*zip(*[c1,c2]*10))
row = ClassEnvironmentTask().process(simulation,simulation.read())
json.dumps(row)
def test_classification_statistics_dense_sans_sklearn(self):
with unittest.mock.patch('importlib.import_module', side_effect=ImportError()):
simulation = SupervisedSimulation([[1,2],[3,4]]*10,["A","B"]*10)
row = ClassEnvironmentTask().process(simulation,simulation.read())
self.assertEqual(2, row["action_cardinality"])
self.assertEqual(2, row["context_dimensions"])
self.assertEqual(1, row["imbalance_ratio"])
self.assertNotIn("bayes_rate_avg",row)
self.assertNotIn("bayes_rate_iqr",row)
self.assertNotIn("centroid_purity",row)
self.assertNotIn("centroid_distance",row)
def test_classification_statistics_dense(self):
env = SupervisedSimulation([[1,2],[3,4]]*10,["A","B"]*10)
row = ClassEnvironmentTask().process(env,env.read())
self.assertEqual(2, row["action_cardinality"])
self.assertEqual(2, row["context_dimensions"])
self.assertEqual(1, row["imbalance_ratio"])
self.assertEqual(1, row["bayes_rate_avg"])
self.assertEqual(0, row["bayes_rate_iqr"])
self.assertEqual(1, row["centroid_purity"])
self.assertEqual(0, row["centroid_distance"])
def test_classification_statistics_sparse_sans_sklearn(self):
with unittest.mock.patch('importlib.import_module', side_effect=ImportError()):
c1 = [{"1":1, "2":2}, "A"]
c2 = [{"1":3, "2":4}, "B"]
simulation = SupervisedSimulation(*zip(*[c1,c2]*10))
row = ClassEnvironmentTask().process(simulation,simulation.read())
self.assertEqual(2, row["action_cardinality"])
self.assertEqual(2, row["context_dimensions"])
self.assertEqual(1, row["imbalance_ratio"])
self.assertNotIn("bayes_rate_avg",row)
self.assertNotIn("bayes_rate_iqr",row)
self.assertNotIn("centroid_purity",row)
self.assertNotIn("centroid_distance",row)
def test_classification_statistics_sparse(self):
c1 = [{"1":1, "2":2}, "A"]
c2 = [{"1":3, "2":4}, "B"]
env = SupervisedSimulation(*zip(*[c1,c2]*10))
row = ClassEnvironmentTask().process(env,env.read())
self.assertEqual(2, row["action_cardinality"])
self.assertEqual(2, row["context_dimensions"])
self.assertEqual(1, row["imbalance_ratio"])
self.assertEqual(1, row["bayes_rate_avg"])
self.assertEqual(0, row["bayes_rate_iqr"])
self.assertEqual(1, row["centroid_purity"])
self.assertEqual(0, row["centroid_distance"])
def test_X_Y_empty(self):
features = []
labels = []
interactions = list(SupervisedSimulation(features, labels).read())
self.assertEqual(len(interactions), 0)
def test_X_Y_multilabel_classification(self):
features = [(8.1, 27, 1410, (0, 1)), (8.2, 29, 1180, (0, 1)),
(8.3, 27, 1020, (1, 0))]
labels = [[1, 2, 3, 4], [1, 2], [1]]
interactions = list(SupervisedSimulation(features, labels).read())
self.assertEqual(len(interactions), 3)
for rnd in interactions:
hash(rnd.context) #make sure these are hashable
hash(rnd.actions[0]) #make sure these are hashable
hash(rnd.actions[1]) #make sure these are hashable
self.assertEqual((8.1, 27, 1410, (0, 1)), interactions[0].context)
self.assertEqual((8.2, 29, 1180, (0, 1)), interactions[1].context)
self.assertEqual((8.3, 27, 1020, (1, 0)), interactions[2].context)
self.assertEqual([1, 4, 3, 2], interactions[0].actions)
self.assertEqual([1, 4, 3, 2], interactions[1].actions)
self.assertEqual([1, 4, 3, 2], interactions[2].actions)
self.assertEqual([1, 1, 1, 1], interactions[0].kwargs["rewards"])
self.assertEqual([1, 0, 0, 1], interactions[1].kwargs["rewards"])
self.assertEqual([1, 0, 0, 0], interactions[2].kwargs["rewards"])
def test_params(self):
expected_params = {
'source': "[X,Y]",
'label_type': 'C',
"type": "SupervisedSimulation"
}
self.assertEqual(expected_params,
SupervisedSimulation([1, 2], [1, 2]).params)
def test_X_Y_too_large_take(self):
features = [(8.1, 27, 1410, (0, 1)), (8.2, 29, 1180, (0, 1)),
(8.3, 27, 1020, (1, 0))]
labels = [2, 2, 1]
interactions = list(
SupervisedSimulation(features, labels, take=4).read())
self.assertEqual(len(interactions), 3)
def test_X_Y_regression_more_than_10(self):
features = list(range(12))
labels = list(range(12))
interactions = list(
SupervisedSimulation(features, labels, label_type='R').read())
self.assertEqual(len(interactions), 12)
self.assertEqual(0, interactions[0].context)
self.assertEqual(1, interactions[1].context)
self.assertEqual(2, interactions[2].context)
self.assertEqual(3, interactions[3].context)
self.assertEqual(4, interactions[4].context)
self.assertEqual(5, interactions[5].context)
self.assertEqual(6, interactions[6].context)
self.assertEqual(7, interactions[7].context)
self.assertEqual(8, interactions[8].context)
self.assertEqual(9, interactions[9].context)
self.assertEqual(10, interactions[10].context)
self.assertEqual(11, interactions[11].context)
actions = [(0, 0, 0, 0, 0, 0, 0, 0, 1, 0),
(0, 1, 0, 0, 0, 0, 0, 0, 0, 0),
(0, 0, 0, 0, 1, 0, 0, 0, 0, 0),
(0, 0, 0, 0, 0, 1, 0, 0, 0, 0),
(0, 0, 0, 1, 0, 0, 0, 0, 0, 0),
(0, 0, 0, 0, 0, 0, 0, 0, 0, 1),
(0, 0, 1, 0, 0, 0, 0, 0, 0, 0),
(0, 0, 0, 0, 0, 0, 1, 0, 0, 0),
(0, 0, 0, 0, 0, 0, 0, 1, 0, 0),
(1, 0, 0, 0, 0, 0, 0, 0, 0, 0)]
self.assertEqual(actions, interactions[0].actions)
self.assertEqual(actions, interactions[1].actions)
self.assertEqual(actions, interactions[2].actions)
self.assertEqual(actions, interactions[3].actions)
self.assertEqual(actions, interactions[4].actions)
self.assertEqual(actions, interactions[5].actions)
self.assertEqual(actions, interactions[6].actions)
self.assertEqual(actions, interactions[7].actions)
self.assertEqual(actions, interactions[8].actions)
self.assertEqual(actions, interactions[9].actions)
self.assertEqual(actions, interactions[10].actions)
self.assertEqual(actions, interactions[11].actions)
for i in range(12):
self.assertEqual([
1 - (abs((a.index(1) + 1) / 11 - i / 11))
for a in interactions[i].actions
], interactions[i].kwargs["rewards"])
def test_source_reader_regression_less_than_10(self):
source = ArffSource(
ListSource("""
@relation weather
@attribute pH real
@attribute temperature real
@attribute conductivity real
@attribute coli {2, 1}
@attribute play {yes, no}
@data
8.1,27,1410,2,no
8.2,29,1180,2,no
8.3,27,1020,1,yes
""".splitlines()))
interactions = list(
SupervisedSimulation(source, "pH", label_type="R").read())
self.assertEqual(len(interactions), 3)
for rnd in interactions:
hash(rnd.context) #make sure these are hashable
hash(rnd.actions[0]) #make sure these are hashable
hash(rnd.actions[1]) #make sure these are hashable
self.assertEqual((27, 1410, (1, 0), (0, 1)), interactions[0].context)
self.assertEqual((29, 1180, (1, 0), (0, 1)), interactions[1].context)
self.assertEqual((27, 1020, (0, 1), (1, 0)), interactions[2].context)
self.assertEqual([(0, 0, 1), (1, 0, 0), (0, 1, 0)],
interactions[0].actions)
self.assertEqual([(0, 0, 1), (1, 0, 0), (0, 1, 0)],
interactions[1].actions)
self.assertEqual([(0, 0, 1), (1, 0, 0), (0, 1, 0)],
interactions[2].actions)
self.assertEqual([0, 1, .5], interactions[0].kwargs["rewards"])
self.assertEqual([.5, .5, 1], interactions[1].kwargs["rewards"])
self.assertEqual([1, 0, .5], interactions[2].kwargs["rewards"])
def test_source_reader_too_large_take_exact_min(self):
source = ArffSource(
ListSource("""
@relation weather
@attribute pH real
@attribute temperature real
@attribute conductivity real
@attribute coli {2, 1}
@attribute play {yes, no}
@data
8.1,27,1410,2,no
8.2,29,1180,2,no
8.3,27,1020,1,yes
""".splitlines()))
interactions = list(
SupervisedSimulation(source, "coli", take=(5, 5)).read())
self.assertEqual(len(interactions), 0)
def test_environment_pipe_statistics_dense(self):
env = Pipes.join(SupervisedSimulation([[1,2],[3,4]]*10,["A","B"]*10), Shuffle(1))
task = SimpleEnvironmentTask()
self.assertEqual({**env.params}, task.process(env,env.read()))
def test_classification_statistics_dense(self):
env = SupervisedSimulation([[1,2],[3,4]]*10,["A","B"]*10)
task = SimpleEnvironmentTask()
self.assertEqual({**env.params}, task.process(env,env.read()))