class TestExperimentLogisticRegression(unittest.TestCase):
def test_run_and_analyze(self):
self.lr16_4 = ExperimentLogisticRegression('exp1.log', 8, 2, 2**8,
0xbeef, 0xbeef,
LTFArray.transform_id,
LTFArray.combiner_xor)
self.lr16_4.execute()
def test_fix_convergence(self, logger):
"""
This methods checks the results of a ExperimentLogisticRegression to match a recorded result which was
generated with the same parameters.
"""
n = 8
k = 2
N = 255
seed_instance = 0xBAE55E
seed_model = 0x5C6AE1E
seed_challenge = 0xB0661E
seed_distance = 0xB0C
experiment = ExperimentLogisticRegression(
LOG_PATH + 'exp',
LRParameters(n=n,
k=k,
N=N,
seed_model=seed_model,
seed_distance=seed_distance,
seed_instance=seed_instance,
seed_challenge=seed_challenge,
transformation='soelter_lightweight_secure',
combiner='xor',
mini_batch_size=0,
shuffle=False,
convergence_decimals=2))
experiment.execute(logger.queue, logger.logger_name)
result_str = logger.read_result_log()
self.assertFalse(result_str == '', 'The result log was empty.')
error = 'LR learning results deviate from legacy learning results.'
self.assertEqual(experiment.result.iteration_count, 274, error)
self.assertEqual(experiment.result.epoch_count, 274, error)
self.assertEqual(experiment.result.gradient_step_count, 274, error)
self.assertEqual(experiment.result.accuracy, 0.98828125, error)
print(experiment.result.model)
assert_array_equal(
around(experiment.result.model, decimals=8),
around([
0.00351544,
-0.00504143,
0.01470355,
0.00481524,
0.00649157,
-0.00301955,
-0.0025765,
0.00841928,
0.,
0.03248558,
0.37524702,
-0.0683109,
0.40447738,
0.49995907,
0.52796785,
0.00060493,
-0.40093716,
0.,
],
decimals=8), error)
def test_run_and_analyze(self, logger):
"""
This method only runs the experiment.
"""
lr16_4 = ExperimentLogisticRegression(
LOG_PATH + 'exp1',
8,
2,
2**8,
0xbeef,
0xbeef,
LTFArray.transform_id,
LTFArray.combiner_xor,
)
lr16_4.execute(logger.queue, logger.logger_name)
def test_fix_convergence(self, logger):
"""
This methods checks the results of a ExperimentLogisticRegression to match a recorded result which was
generated with the same parameters.
"""
n = 8
k = 2
N = 255
seed_instance = 0xBAE55E
seed_model = 0x5C6AE1E
seed_challenge = 0xB0661E
seed_distance = 0xB0C
experiment = ExperimentLogisticRegression(
LOG_PATH + 'exp',
LRParameters(n=n,
k=k,
N=N,
seed_model=seed_model,
seed_distance=seed_distance,
seed_instance=seed_instance,
seed_challenge=seed_challenge,
transformation='soelter_lightweight_secure',
combiner='xor',
mini_batch_size=0,
shuffle=False,
convergence_decimals=2))
experiment.execute(logger.queue, logger.logger_name)
result_str = logger.read_result_log()
self.assertFalse(result_str == '', 'The result log was empty.')
error = 'LR learning results deviate from legacy learning results.'
self.assertEqual(experiment.result.iteration_count, 256, error)
self.assertEqual(experiment.result.epoch_count, 256, error)
self.assertEqual(experiment.result.gradient_step_count, 256, error)
self.assertEqual(experiment.result.accuracy, 0.98828125, error)
assert_array_equal(
around(experiment.result.model, decimals=8),
around([
3.99071615e-03, -5.65532862e-03, 1.63862406e-02,
5.37762262e-03, 7.29781422e-03, -3.35141930e-03,
-2.95642974e-03, 9.40114614e-03, 1.26572532e-07,
3.49183531e-02, 3.68758330e-01, -7.85028286e-02,
4.17595994e-01, 5.09973673e-01, 5.13855116e-01, 2.97216086e-04,
-3.96978992e-01, -5.41390228e-03
],
decimals=8), error)
def test_run_and_analyze(self, logger):
"""
This method only runs the experiment.
"""
lr16_4 = ExperimentLogisticRegression(
LOG_PATH + 'exp1',
LRParameters(n=8,
k=2,
N=2**8,
seed_model=0xbeef,
seed_distance=0xbeef,
seed_instance=0xdead,
seed_challenge=0xdead,
transformation='id',
combiner='xor',
mini_batch_size=2,
shuffle=False,
convergence_decimals=2))
lr16_4.execute(logger.queue, logger.logger_name)
def test_fix_convergence(self, logger):
"""
This methods checks the results of a ExperimentLogisticRegression to match a recorded result which was
generated with the same parameters.
"""
n = 8
k = 2
N = 255
seed_instance = 0xBAE55E
seed_model = 0x5C6AE1E
seed_challenge = 0xB0661E
seed_distance = 0xB0C
experiment = ExperimentLogisticRegression(
LOG_PATH + 'exp',
n,
k,
N,
seed_instance,
seed_model,
LTFArray.transform_soelter_lightweight_secure,
LTFArray.combiner_xor,
seed_challenge=seed_challenge,
seed_chl_distance=seed_distance,
)
experiment.execute(logger.queue, logger.logger_name)
legacy_result = [
'0xbae55e', '0x5c6ae1e', '0', '8', '2', '255',
'transform_soelter_lightweight_secure', 'combiner_xor', '363',
'1.000000', '0.00443419669755,-0.00616546911566,0.0186346081194,'
'0.0061619719475,0.00795284461334,-0.00443539877583,'
'-0.00316047872599,0.00993214368373,0.0507595729459,'
'0.415207373134,-0.0517173737839,0.285900582842,'
'0.467512016377,0.550102231366,-0.000739711610042,'
'-0.467757977178\n'
]
result_str = logger.read_result_log()
self.assertFalse(result_str == '', 'The result log was empty.')
experiment_result = result_str.split('\t')
# remove execution time
del experiment_result[9]
self.assertTrue(experiment_result == legacy_result,
'You changed the code significant.')
class TestExperimentLogisticRegression(TestBase):
def test_run_and_analyze(self):
logger_name = 'log'
# Setup multiprocessing logging
queue = multiprocessing.Queue(-1)
listener = multiprocessing.Process(target=log_listener,
args=(
queue,
setup_logger,
logger_name,
))
listener.start()
self.lr16_4 = ExperimentLogisticRegression('exp1', 8, 2, 2**8, 0xbeef,
0xbeef,
LTFArray.transform_id,
LTFArray.combiner_xor)
self.lr16_4.execute(queue, logger_name)
queue.put_nowait(None)
listener.join()