def wrapper_grad_posterior_mean_bq(point,
self,
var_noise=None,
mean=None,
parameters_kernel=None,
n_samples_parameters=0):
"""
Wrapper of the gradient of the posterior mean of a bq model
:param point: np.array(k)
:param self: bayesian-quadrature instance
:param n_samples_parameters: int
:return: np.array(k)
"""
if n_samples_parameters == 0:
val = self.grad_posterior_mean(point,
var_noise=var_noise,
mean=mean,
parameters_kernel=parameters_kernel)
else:
val = BayesianEvaluations.evaluate(self.grad_posterior_mean, point,
self.gp, n_samples_parameters,
None)[0]
return val
def wrapper_objective_acquisition_function(point,
self,
n_samples_parameters=0,
*params):
"""
Wrapper of an acquisition function that's not SBO or KG.
:param point: np.array(n)
:param self: acquisition function instance
:param n_samples_parameters: int
:param params: additional parameters of the function
:return: float
"""
point = point.reshape((1, len(point)))
if n_samples_parameters == 0:
value = self.evaluate(point, *params)
else:
if self.gp.name_model == BAYESIAN_QUADRATURE:
gp_model = self.gp.gp
else:
gp_model = self.gp
value = BayesianEvaluations.evaluate(self.evaluate, point, gp_model,
n_samples_parameters, None,
*params)[0]
return value
def wrapper_gradient_posterior_mean_gp_model(point,
self,
n_samples_parameters=0,
*params):
point = point.reshape((1, len(point)))
if n_samples_parameters == 0:
value = self.gradient_posterior_parameters(point,
*params,
only_mean=True)['mean']
else:
def evaluate(point, var_noise=None, mean=None, parameters_kernel=None):
return self.gradient_posterior_parameters(point,
var_noise,
mean,
parameters_kernel,
only_mean=True)['mean']
value = BayesianEvaluations.evaluate(evaluate, point, self,
n_samples_parameters, None,
*params)[0]
return value