def it_requests_sampling_when_delta_is_in_interval(self):
class FakeCI:
def two_sided_interval(self, matrix, profile, alpha):
return [0.0, 1.0]
ci_calculator = FakeCI()
matrix = create_observation_matrix()
evaluator = ConfidenceIntervalEvaluator(matrix.toGame(), matrix.knownProfiles(), 0.5, 0.95, ci_calculator)
assert evaluator.continue_sampling(matrix) == True
def it_tracks_whether_or_not_a_profile_is_a_delta_nash(self):
class FakeCI:
def __init__(self):
self.count = 0
def two_sided_interval(self, matrix, profile, alpha):
self.count += 1
if self.count % 2 == 0:
return [0.6, 1.0]
else:
return [0.0, 0.4]
ci_calculator = FakeCI()
matrix = create_observation_matrix()
evaluator = ConfidenceIntervalEvaluator(matrix.toGame(), matrix.knownProfiles(), 0.5, 0.95, ci_calculator)
evaluator.continue_sampling(matrix)
count = 0
print evaluator.target_hash.items()
for result in evaluator.target_hash.values():
count += 1
if count % 2 == 0:
assert result == "No"
else:
assert result == "Yes"
def it_does_not_request_sampling_when_delta_is_outside_all_intervals(self):
class FakeCI:
def __init__(self):
self.count = 0
def two_sided_interval(self, matrix, profile, alpha):
self.count += 1
if self.count % 2 == 0:
return [0.6, 1.0]
else:
return [0.0, 0.4]
ci_calculator = FakeCI()
matrix = create_observation_matrix()
evaluator = ConfidenceIntervalEvaluator(matrix.toGame(), matrix.knownProfiles(), 0.5, 0.95, ci_calculator)
assert evaluator.continue_sampling(matrix) == False
def it_requests_sampling_when_given_an_empty_observation_matrix(self):
matrix = ObservationMatrix()
evaluator = ConfidenceIntervalEvaluator(create_symmetric_game(), [], 0.5, 0.95, None)
assert evaluator.continue_sampling(matrix) == True