def it_can_add_observations_to_existing_profiles(self):
matrix = ObservationMatrix([{'All': [PayoffData('A', 2, [12, 11])]},
{'All': [PayoffData('A', 1, [23]), PayoffData('B', 1, [11])]},
{'All': [PayoffData('B', 2, [13.3])]}])
matrix.addObservations(Profile({'All': {'A': 1, 'B': 1}}),
{'All': [PayoffData('A', 1, [12, 11]), PayoffData('B', 1, [21, 17])]})
assert matrix.getPayoffData(Profile({'All': {'A': 1, 'B': 1}}), 'All', 'B') == [11, 21, 17]
def it_does_not_request_sampling_when_only_profiles_outside_target_set_have_too_much_variation_in_payoffs(self):
matrix = ObservationMatrix()
target_profile = Profile({'All': {'A': 2}})
non_target_profile = Profile({'All': {'B': 2}})
matrix.addObservations(target_profile, {'All': [PayoffData('A', 2, [10, 10.1, 10.2])]})
matrix.addObservations(non_target_profile, {'All': [PayoffData('B', 2, [10, 20, 45])]})
evaluator = StandardErrorEvaluator(1, [target_profile])
assert evaluator.continue_sampling(matrix) == False
def add_noise_sequentially(game, model, evaluator, samples_per_step):
"""
Generate ObservationMatrix sequentially with bimodal gaussian noise added to each payoff.
game: a RSG.Game or RSG.SampleGame
model: noise model
samples_per_step: the parameters passed to the noise function
evaluator: the method used to determine when to stop sampling
"""
matrix = ObservationMatrix()
count = 0
while evaluator.continue_sampling(matrix) and count < 1000:
for prof in game.knownProfiles():
matrix.addObservations(prof, model.generate_samples(game, prof, samples_per_step))
count += samples_per_step
print count
return [matrix, evaluator.equilibria()]
def it_can_add_observations_for_previously_unobserved_profiles(self):
matrix = ObservationMatrix()
matrix.addObservations(Profile({'All': {'A': 1, 'B': 1}}),
{'All': [PayoffData('A', 1, [12, 11]), PayoffData('B', 1, [21, 17])]})
assert matrix.getPayoffData(Profile({'All': {'A': 1, 'B': 1}}), 'All', 'B') == [21, 17]
