def test_lower_must_be_less_than_upper(self):
with raises(ValueError):
Bound(lower=1.0, upper=0.0)
# Now we'll set up the main BO object. First we need to create a surrogate model.
# We'll use a GPyGPSurrogate with an RBF kernel. See the GPyGPSurrogate class for
# why we have to create this slightly funky initializer function first.
def gp_initializer(x, y):
return GPy.models.GPRegression(x,
y,
kernel=GPy.kern.RBF(input_dim=1),
noise_var=1e-10,
normalizer=True)
surrogate = GPyGPSurrogate(gp_initializer=gp_initializer)
acquistion_function = LCB(surrogate=surrogate)
bounds = Bounds(bounds=[Bound(lower=0.0, upper=1.0)])
optimizer = DirectOptimizer(acquisition_function=acquistion_function,
bounds=bounds,
maxf=100)
# Now we create the BO object...
bo = BayesOpt(
objective_function=forrester,
surrogate=surrogate,
acquisition_function=acquistion_function,
optimizer=optimizer,
bounds=bounds,
initial_design=SobolSequenceInitialDesign(),
callbacks=[PlottingCallback()],
)
def bounds(self, lowers, uppers):
return Bounds(
bounds=[Bound(lower=l, upper=u) for l, u in zip(lowers, uppers)])
def bounds():
return Bounds(bounds=[Bound(lower=0.0, upper=1.0)])