def test_optimize_normalize_Y():
np.random.seed(42)
arr_range = np.array([
[0.0, 10.0],
[-2.0, 2.0],
[-5.0, 5.0],
])
dim_X = arr_range.shape[0]
num_X = 5
X = np.random.randn(num_X, dim_X)
Y = np.random.randn(num_X, 1)
model_bo = BO(arr_range, normalize_Y=True)
next_point, dict_info = model_bo.optimize(X, Y)
Y_original = dict_info['Y_original']
Y_normalized = dict_info['Y_normalized']
assert np.all(Y == Y_original)
assert np.all(Y != Y_normalized)
assert np.all(utils_bo.normalize_min_max(Y) == Y_normalized)
model_bo = BO(arr_range, normalize_Y=False)
next_point, dict_info = model_bo.optimize(X, Y)
Y_original = dict_info['Y_original']
Y_normalized = dict_info['Y_normalized']
assert np.all(Y == Y_normalized)
assert np.all(Y == Y_original)
def test_normalize_min_max():
Y = np.array([
[1.0],
[2.0],
[10.0],
[-5.0],
[4.0],
[2.0],
[-4.0],
[2.0],
])
with pytest.raises(AssertionError) as error:
package_target.normalize_min_max(123)
with pytest.raises(AssertionError) as error:
package_target.normalize_min_max('abc')
with pytest.raises(AssertionError) as error:
package_target.normalize_min_max(np.squeeze(Y))
Y_normalized = package_target.normalize_min_max(Y)
truth_Y = np.array([
[6.0 / 15.0],
[7.0 / 15.0],
[1.0],
[0.0],
[9.0 / 15.0],
[7.0 / 15.0],
[1.0 / 15.0],
[7.0 / 15.0],
]) * constants.MULTIPLIER_RESPONSE
assert np.all(Y_normalized == truth_Y)
Y = np.array([
[10.0],
[10.0],
[10.0],
])
Y_normalized = package_target.normalize_min_max(Y)
assert np.all(Y_normalized == Y)
def thompson_sampling_gp_iteration(range_X: np.ndarray,
X: np.ndarray, Y: np.ndarray,
normalize_Y: bool=constants.NORMALIZE_RESPONSE,
str_sampling_method: str='sobol',
num_samples: int=200,
debug: bool=False,
) -> np.ndarray:
"""
It chooses the next query point via Thompson sampling.
:param range_X: bounds for a search space. Shape: (d, 2).
:type range_X: numpy.ndarray
:param X: inputs. Shape: (n, d).
:type X: numpy.ndarray
:param Y: outputs. Shape: (n, 1).
:type Y: numpy.ndarray
:param normalize_Y: flag for normalizing responses.
:type normalize_Y: bool., optional
:param str_sampling_method: the name of sampling method.
:type str_sampling_method: str., optional
:param num_samples: the number of samples.
:type num_samples: int., optional
:param debug: flag for a debug option.
:type debug: bool., optional
:returns: the next point. Shape: (d, ).
:rtype: numpy.ndarray
:raises: AssertionError, ValueError
"""
assert isinstance(range_X, np.ndarray)
assert isinstance(X, np.ndarray)
assert isinstance(Y, np.ndarray)
assert isinstance(normalize_Y, bool)
assert isinstance(str_sampling_method, str)
assert isinstance(num_samples, int)
assert isinstance(debug, bool)
assert len(range_X.shape) == 2
assert len(X.shape) == 2
assert len(Y.shape) == 2
assert X.shape[0] == Y.shape[0]
assert range_X.shape[0] == X.shape[1]
assert range_X.shape[1] == 2
str_cov = 'matern52'
prior_mu = None
str_optimizer_method_gp = 'BFGS'
# use_ard = True
if normalize_Y:
if debug:
logger.debug('Responses are normalized.')
Y = utils_bo.normalize_min_max(Y)
model_bo = bo.BOwGP(range_X)
X_test = model_bo.get_samples(str_sampling_method, num_samples=num_samples)
mu_Xs, _, Sigma_Xs = gp.predict_with_optimized_hyps(X, Y, X_test,
str_cov=str_cov, str_optimizer_method=str_optimizer_method_gp,
prior_mu=prior_mu, debug=debug)
mu_Xs = np.squeeze(mu_Xs, axis=1)
Y_sampled = None
list_jitters = [0.0, 1e-4, 1e-2, 1e0, 1e1, 1e2, 1e3, 1e4]
for jitter_cov in list_jitters:
try:
Sigma_Xs_ = Sigma_Xs + jitter_cov * np.eye(Sigma_Xs.shape[0])
Y_sampled = gp.sample_functions(mu_Xs, Sigma_Xs_, num_samples=1)
break
except ValueError: # pragma: no cover
pass
if Y_sampled is None: # pragma: no cover
raise ValueError('jitter_cov is not large enough.')
ind_min = np.argmin(Y_sampled[:, 0])
next_point = X_test[ind_min]
return next_point
def optimize(
self,
X_train: np.ndarray,
Y_train: np.ndarray,
str_sampling_method: str = constants.STR_SAMPLING_METHOD_AO,
num_samples: int = constants.NUM_SAMPLES_AO,
str_mlm_method: str = constants.STR_MLM_METHOD,
) -> constants.TYPING_TUPLE_ARRAY_DICT:
"""
It computes acquired example, candidates of acquired examples,
acquisition function values for the candidates, covariance matrix,
inverse matrix of the covariance matrix, hyperparameters optimized,
and execution times.
:param X_train: inputs. Shape: (n, d) or (n, m, d).
:type X_train: numpy.ndarray
:param Y_train: outputs. Shape: (n, 1).
:type Y_train: numpy.ndarray
:param str_sampling_method: the name of sampling method for
acquisition function optimization.
:type str_sampling_method: str., optional
:param num_samples: the number of samples.
:type num_samples: int., optional
:param str_mlm_method: the name of marginal likelihood maximization
method for Gaussian process regression.
:type str_mlm_method: str., optional
:returns: acquired example and dictionary of information. Shape: ((d, ), dict.).
:rtype: (numpy.ndarray, dict.)
:raises: AssertionError
"""
assert isinstance(X_train, np.ndarray)
assert isinstance(Y_train, np.ndarray)
assert isinstance(str_sampling_method, str)
assert isinstance(num_samples, int)
assert isinstance(str_mlm_method, str)
assert len(X_train.shape) == 2
assert len(Y_train.shape) == 2
assert Y_train.shape[1] == 1
assert X_train.shape[0] == Y_train.shape[0]
assert X_train.shape[1] == self.num_dim
assert num_samples > 0
assert str_sampling_method in constants.ALLOWED_SAMPLING_METHOD
assert str_mlm_method in constants.ALLOWED_MLM_METHOD
time_start = time.time()
Y_train_orig = Y_train
if self.normalize_Y and str_mlm_method != 'converged':
if self.debug:
self.logger.debug('Responses are normalized.')
Y_train = utils_bo.normalize_min_max(Y_train)
time_start_surrogate = time.time()
if str_mlm_method == 'regular':
cov_X_X, inv_cov_X_X, hyps = gp_kernel.get_optimized_kernel(
X_train,
Y_train,
self.prior_mu,
self.str_cov,
str_optimizer_method=self.str_optimizer_method_gp,
str_modelselection_method=self.str_modelselection_method,
use_ard=self.use_ard,
debug=self.debug)
elif str_mlm_method == 'combined':
from bayeso.gp import gp_likelihood
from bayeso.utils import utils_gp
from bayeso.utils import utils_covariance
prior_mu_train = utils_gp.get_prior_mu(self.prior_mu, X_train)
neg_log_ml_best = np.inf
cov_X_X_best = None
inv_cov_X_X_best = None
hyps_best = None
for cur_str_optimizer_method in ['BFGS', 'Nelder-Mead']:
cov_X_X, inv_cov_X_X, hyps = gp_kernel.get_optimized_kernel(
X_train,
Y_train,
self.prior_mu,
self.str_cov,
str_optimizer_method=cur_str_optimizer_method,
str_modelselection_method=self.str_modelselection_method,
use_ard=self.use_ard,
debug=self.debug)
cur_neg_log_ml_ = gp_likelihood.neg_log_ml(
X_train,
Y_train,
utils_covariance.convert_hyps(
self.str_cov, hyps, fix_noise=constants.FIX_GP_NOISE),
self.str_cov,
prior_mu_train,
use_ard=self.use_ard,
fix_noise=constants.FIX_GP_NOISE,
use_gradient=False,
debug=self.debug)
if cur_neg_log_ml_ < neg_log_ml_best:
neg_log_ml_best = cur_neg_log_ml_
cov_X_X_best = cov_X_X
inv_cov_X_X_best = inv_cov_X_X
hyps_best = hyps
cov_X_X = cov_X_X_best
inv_cov_X_X = inv_cov_X_X_best
hyps = hyps_best
elif str_mlm_method == 'converged':
fix_noise = constants.FIX_GP_NOISE
if self.is_optimize_hyps:
cov_X_X, inv_cov_X_X, hyps = gp_kernel.get_optimized_kernel(
X_train,
Y_train,
self.prior_mu,
self.str_cov,
str_optimizer_method=self.str_optimizer_method_gp,
str_modelselection_method=self.str_modelselection_method,
use_ard=self.use_ard,
debug=self.debug)
self.is_optimize_hyps = not utils_bo.check_hyps_convergence(
self.historical_hyps, hyps, self.str_cov, fix_noise)
else: # pragma: no cover
if self.debug:
self.logger.debug('hyps converged.')
hyps = self.historical_hyps[-1]
cov_X_X, inv_cov_X_X, _ = covariance.get_kernel_inverse(
X_train,
hyps,
self.str_cov,
fix_noise=fix_noise,
debug=self.debug)
else: # pragma: no cover
raise ValueError('optimize: missing condition for str_mlm_method.')
self.historical_hyps.append(hyps)
time_end_surrogate = time.time()
time_start_acq = time.time()
fun_negative_acquisition = lambda X_test: -1.0 * self.compute_acquisitions(
X_test, X_train, Y_train, cov_X_X, inv_cov_X_X, hyps)
next_point, next_points = self._optimize(
fun_negative_acquisition,
str_sampling_method=str_sampling_method,
num_samples=num_samples)
time_end_acq = time.time()
acquisitions = fun_negative_acquisition(next_points)
time_end = time.time()
dict_info = {
'next_points': next_points,
'acquisitions': acquisitions,
'Y_original': Y_train_orig,
'Y_normalized': Y_train,
'cov_X_X': cov_X_X,
'inv_cov_X_X': inv_cov_X_X,
'hyps': hyps,
'time_surrogate': time_end_surrogate - time_start_surrogate,
'time_acq': time_end_acq - time_start_acq,
'time_overall': time_end - time_start,
}
if self.debug:
self.logger.debug('overall time consumed to acquire: %.4f sec.',
time_end - time_start)
return next_point, dict_info
def optimize(
self,
X_train: np.ndarray,
Y_train: np.ndarray,
str_sampling_method: str = constants.STR_SAMPLING_METHOD_AO_TREES,
num_samples: int = constants.NUM_SAMPLES_AO_TREES,
) -> constants.TYPING_TUPLE_ARRAY_DICT:
"""
It computes acquired example, candidates of acquired examples,
acquisition function values for the candidates, covariance matrix,
inverse matrix of the covariance matrix, hyperparameters optimized,
and execution times.
:param X_train: inputs. Shape: (n, d).
:type X_train: numpy.ndarray
:param Y_train: outputs. Shape: (n, 1).
:type Y_train: numpy.ndarray
:param str_sampling_method: the name of sampling method for
acquisition function optimization.
:type str_sampling_method: str., optional
:param num_samples: the number of samples.
:type num_samples: int., optional
:returns: acquired example and dictionary of information. Shape: ((d, ), dict.).
:rtype: (numpy.ndarray, dict.)
:raises: AssertionError
"""
assert isinstance(X_train, np.ndarray)
assert isinstance(Y_train, np.ndarray)
assert isinstance(str_sampling_method, str)
assert isinstance(num_samples, int)
assert len(X_train.shape) == 2
assert len(Y_train.shape) == 2
assert Y_train.shape[1] == 1
assert X_train.shape[0] == Y_train.shape[0]
assert X_train.shape[1] == self.num_dim
assert num_samples > 0
assert str_sampling_method in constants.ALLOWED_SAMPLING_METHOD
time_start = time.time()
Y_train_orig = Y_train
if self.normalize_Y:
if self.debug:
self.logger.debug('Responses are normalized.')
Y_train = utils_bo.normalize_min_max(Y_train)
time_start_surrogate = time.time()
trees = self.get_trees(X_train, Y_train)
time_end_surrogate = time.time()
time_start_acq = time.time()
next_points = self.get_samples(str_sampling_method,
fun_objective=None,
num_samples=num_samples)
fun_negative_acquisition = lambda X_test: -1.0 * self.compute_acquisitions(
X_test, X_train, Y_train, trees)
acquisitions = fun_negative_acquisition(next_points)
ind_next_point = np.argmin(acquisitions)
next_point = next_points[ind_next_point]
time_end_acq = time.time()
time_end = time.time()
dict_info = {
'next_points': next_points,
'acquisitions': acquisitions,
'Y_original': Y_train_orig,
'Y_normalized': Y_train,
'trees': trees,
'time_surrogate': time_end_surrogate - time_start_surrogate,
'time_acq': time_end_acq - time_start_acq,
'time_overall': time_end - time_start,
}
if self.debug:
self.logger.debug('overall time consumed to acquire: %.4f sec.',
time_end - time_start)
return next_point, dict_info