def test_mutate(self):
protocol = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=1.0, duration=0.5),
protocols.VoltageClampStep(voltage=2.0, duration=0.75),
])
# Finds seed so that only second step is mutated.
for i in range(500):
random.seed(i)
num_one = random.random()
num_two = random.random()
if num_two < self.vc_ga.config.gene_mutation_probability < num_one:
random.seed(i)
break
else:
self.fail(msg='Could not find seed to meet required behavior.')
np.random.seed(3)
self.vc_ga._mutate(
individual=genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol))
expected_protocol = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=1.0, duration=0.5),
protocols.VoltageClampStep(voltage=0.6,
duration=1.0264562386575933),
])
self.assertEqual(protocol, expected_protocol)
def test_evaluate_returns_normal(self):
protocol = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=0.02, duration=0.1),
protocols.VoltageClampStep(voltage=-0.03, duration=0.1865),
])
fitness = self.vc_ga._evaluate(
individual=genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol))
self.assertGreater(fitness, 0)
def _init_individual(self):
"""Initializes a individual with a randomized protocol."""
steps = []
for i in range(self.config.steps_in_protocol):
random_step = protocols.VoltageClampStep(
voltage=random.uniform(*self.config.step_voltage_bounds),
duration=random.uniform(*self.config.step_duration_bounds))
steps.append(random_step)
return genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocols.VoltageClampProtocol(steps=steps), fitness=0)
def test_evaluate_returns_zero_error(self):
protocol = protocols.VoltageClampProtocol(steps=[
# Intentionally set to large voltage to cause null trace.
protocols.VoltageClampStep(voltage=1000.0, duration=0.5),
protocols.VoltageClampStep(voltage=3.7886, duration=1.1865),
])
error = self.vc_ga._evaluate(
individual=genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol))
self.assertEqual(error, 0)
def setUp(self):
vc_ga_config = ga_configs.VoltageOptimizationConfig(
window=100,
step_size=1,
steps_in_protocol=6,
step_duration_bounds=(0.1, 2.0),
step_voltage_bounds=(-1.2, 0.6),
population_size=2,
max_generations=2,
mate_probability=0.9,
mutate_probability=1.0,
gene_swap_probability=0.5,
gene_mutation_probability=0.1,
tournament_size=3)
self.ga_result = genetic_algorithm_results. \
GAResultVoltageClampOptimization(config=vc_ga_config)
sample_vc_protocol = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(duration=0.1, voltage=-0.08),
protocols.VoltageClampStep(duration=0.1, voltage=-0.12),
protocols.VoltageClampStep(duration=0.5, voltage=-0.06),
protocols.VoltageClampStep(duration=0.05, voltage=-0.04),
protocols.VoltageClampStep(duration=0.15, voltage=0.02),
protocols.VoltageClampStep(duration=0.025, voltage=-0.08),
protocols.VoltageClampStep(duration=0.3, voltage=0.04),
])
generation_one = [
genetic_algorithm_results.VCOptimizationIndividual(
protocol=sample_vc_protocol, fitness=0),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=sample_vc_protocol, fitness=2),
]
generation_two = [
genetic_algorithm_results.VCOptimizationIndividual(
protocol=sample_vc_protocol, fitness=5),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=sample_vc_protocol, fitness=9),
]
self.ga_result.generations.append(generation_one)
self.ga_result.generations.append(generation_two)
def test_mate(self):
protocol_one = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=1.0, duration=0.5),
protocols.VoltageClampStep(voltage=2.0, duration=0.75),
])
protocol_two = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=5.0, duration=1.0),
protocols.VoltageClampStep(voltage=-2.0, duration=2.75),
])
for i in range(500):
random.seed(i)
r_num_one = random.random()
r_num_two = random.random()
if r_num_two < self.vc_ga.config.gene_swap_probability < r_num_one:
random.seed(i)
break
else:
self.fail(msg='Could not find seed to meet required behavior.')
self.vc_ga._mate(
i_one=genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol_one),
i_two=genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol_two))
expected_protocol_one = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=1.0, duration=0.5),
protocols.VoltageClampStep(voltage=-2.0, duration=2.75),
])
expected_protocol_two = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=5.0, duration=1.0),
protocols.VoltageClampStep(voltage=2.0, duration=0.75),
])
self.assertEqual(protocol_one, expected_protocol_one)
self.assertEqual(protocol_two, expected_protocol_two)
def test_mutate_steps_falls_between_bounds(self):
protocol = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=-1.0, duration=0.5),
protocols.VoltageClampStep(voltage=0.4, duration=0.75),
])
individual = genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol)
for _ in range(5000):
self.vc_ga._mutate(individual=individual)
for i in individual.protocol.steps:
self.assertTrue(
self.vc_ga.config.step_duration_bounds[0] <= i.duration <=
self.vc_ga.config.step_duration_bounds[1])
self.assertTrue(
self.vc_ga.config.step_voltage_bounds[0] <= i.voltage <=
self.vc_ga.config.step_voltage_bounds[1])
def test_select(self):
protocol = protocols.VoltageClampProtocol(steps=[
protocols.VoltageClampStep(voltage=0.2, duration=0.1),
protocols.VoltageClampStep(voltage=-0.3, duration=0.1865),
])
population = [
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=0),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=5),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=1),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=3.4),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=8),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=10),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=2.2),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=1.2),
]
random.seed(10)
new_population = self.vc_ga._select(population=population)
# Expected population was discovered through printing out
# random.sample() with seed set to 10.
expected_population = [
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=3.4),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=5),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=2.2),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=8),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=1.2),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=10),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=5),
genetic_algorithm_results.VCOptimizationIndividual(
protocol=protocol, fitness=10),
]
self.assertListEqual(new_population, expected_population)