def test_against_MUX(self):
scenario = MUXProblem(training_cycles=10000, address_size=3)
algorithm = xcs.XCSAlgorithm()
algorithm.exploration_probability = .1
algorithm.discount_factor = 0
algorithm.do_ga_subsumption = True
algorithm.do_action_set_subsumption = True
best = None
expected = .6
for _ in range(2):
logging.disable(logging.CRITICAL)
try:
scenario.reset()
steps, total_reward, time_passed, population = xcs.test(
algorithm, scenario)
finally:
logging.disable(logging.NOTSET)
average_reward = total_reward / steps
self.assertGreater(average_reward, .49)
self.assertLess(time_passed, 40)
if average_reward >= expected:
break
elif best is None or best < average_reward:
best = average_reward
else:
self.fail("Failed to achieve expected average reward level. "
"(Missed by %f.)" % (expected - best))
def test_against_MUX(self):
scenario = MUXProblem(training_cycles=10000, address_size=3)
algorithm = xcs.XCSAlgorithm()
algorithm.exploration_probability = .1
algorithm.discount_factor = 0
algorithm.do_ga_subsumption = True
algorithm.do_action_set_subsumption = True
best = None
expected = .6
for _ in range(2):
logging.disable(logging.CRITICAL)
try:
scenario.reset()
steps, total_reward, time_passed, population = xcs.test(
algorithm,
scenario
)
finally:
logging.disable(logging.NOTSET)
average_reward = total_reward / steps
self.assertGreater(average_reward, .49)
self.assertLess(time_passed, 40)
if average_reward >= expected:
break
elif best is None or best < average_reward:
best = average_reward
else:
self.fail("Failed to achieve expected average reward level. "
"(Missed by %f.)" % (expected - best))
def test_using(self):
if self.use_numpy:
self.assertTrue(bitstrings.using_numpy())
self.assertTrue('numpy' in bitstrings.BitString.__module__)
bitstrings.use_pure_python()
self.assertFalse(bitstrings.using_numpy())
self.assertTrue('python' in bitstrings.BitString.__module__)
bitstrings.use_numpy()
self.assertTrue(bitstrings.using_numpy())
self.assertTrue('numpy' in bitstrings.BitString.__module__)
logging.disable(logging.CRITICAL)
try:
xcs.test(scenario=xcs.scenarios.MUXProblem(1000))
finally:
logging.disable(logging.NOTSET)
else:
self.assertFalse(bitstrings.using_numpy())
self.assertTrue('python' in bitstrings.BitString.__module__)
logging.disable(logging.CRITICAL)
try:
xcs.test(scenario=xcs.scenarios.MUXProblem(1000))
finally:
logging.disable(logging.NOTSET)
def test_using(self):
if self.use_numpy:
self.assertTrue(bitstrings.using_numpy())
self.assertTrue('numpy' in bitstrings.BitString.__module__)
bitstrings.use_pure_python()
self.assertFalse(bitstrings.using_numpy())
self.assertTrue('python' in bitstrings.BitString.__module__)
bitstrings.use_numpy()
self.assertTrue(bitstrings.using_numpy())
self.assertTrue('numpy' in bitstrings.BitString.__module__)
logging.disable(logging.CRITICAL)
try:
xcs.test(scenario=xcs.scenarios.MUXProblem(1000))
finally:
logging.disable(logging.NOTSET)
else:
self.assertFalse(bitstrings.using_numpy())
self.assertTrue('python' in bitstrings.BitString.__module__)
logging.disable(logging.CRITICAL)
try:
xcs.test(scenario=xcs.scenarios.MUXProblem(1000))
finally:
logging.disable(logging.NOTSET)
def test_against_haystack(self):
scenario = HaystackProblem(training_cycles=10000, input_size=500)
algorithm = xcs.XCSAlgorithm()
algorithm.ga_threshold = 1
algorithm.crossover_probability = .5
algorithm.exploration_probability = .1
algorithm.discount_factor = 0
algorithm.do_ga_subsumption = False
algorithm.do_action_set_subsumption = True
algorithm.wildcard_probability = 1 - 1 / 500
algorithm.deletion_threshold = 1
algorithm.mutation_probability = 1 / 500
best = None
expected = .6
for _ in range(2):
logging.disable(logging.CRITICAL)
try:
scenario.reset()
steps, total_reward, time_passed, population = xcs.test(
algorithm,
scenario
)
finally:
logging.disable(logging.NOTSET)
average_reward = total_reward / steps
self.assertGreater(average_reward, .48)
self.assertLess(time_passed, 100)
if average_reward >= expected:
break
elif best is None or best < average_reward:
best = average_reward
else:
self.fail("Failed to achieve expected average reward level. "
"(Missed by %f.)" % (expected - best))
def test_against_haystack(self):
scenario = HaystackProblem(training_cycles=10000, input_size=500)
algorithm = xcs.XCSAlgorithm()
algorithm.ga_threshold = 1
algorithm.crossover_probability = .5
algorithm.exploration_probability = .1
algorithm.discount_factor = 0
algorithm.do_ga_subsumption = False
algorithm.do_action_set_subsumption = True
algorithm.wildcard_probability = 1 - 1 / 500
algorithm.deletion_threshold = 1
algorithm.mutation_probability = 1 / 500
best = None
expected = .6
for _ in range(2):
logging.disable(logging.CRITICAL)
try:
scenario.reset()
steps, total_reward, time_passed, population = xcs.test(
algorithm, scenario)
finally:
logging.disable(logging.NOTSET)
average_reward = total_reward / steps
self.assertGreater(average_reward, .48)
self.assertLess(time_passed, 100)
if average_reward >= expected:
break
elif best is None or best < average_reward:
best = average_reward
else:
self.fail("Failed to achieve expected average reward level. "
"(Missed by %f.)" % (expected - best))