def test_constrained_expected_improvement_batch(self):
for dtype in (torch.float, torch.double):
mean = torch.tensor(
[[-0.5, 0.0, 5.0, 0.0], [0.0, 0.0, 5.0, 0.0],
[0.5, 0.0, 5.0, 0.0]],
device=self.device,
dtype=dtype,
).unsqueeze(dim=-2)
variance = torch.ones(3, 4, device=self.device,
dtype=dtype).unsqueeze(dim=-2)
N = torch.distributions.Normal(loc=0.0, scale=1.0)
a = N.icdf(
torch.tensor(0.75)) # get a so that P(-a <= N <= a) = 0.5
mm = MockModel(MockPosterior(mean=mean, variance=variance))
module = ConstrainedExpectedImprovement(
model=mm,
best_f=0.0,
objective_index=0,
constraints={
1: [None, 0],
2: [5.0, None],
3: [-a, a]
},
)
X = torch.empty(3, 1, 1, device=self.device, dtype=dtype) # dummy
ei = module(X)
self.assertTrue(ei.shape == torch.Size([3]))
ei_expected_unconstrained = torch.tensor(
[0.19780, 0.39894, 0.69780], device=self.device, dtype=dtype)
ei_expected = ei_expected_unconstrained * 0.5 * 0.5 * 0.5
self.assertTrue(torch.allclose(ei, ei_expected, atol=1e-4))
def __init__(self, model_list: List[Model], constraints,
options: dict) -> None:
# best_f = torch.min(model_list.models[0].train_targets)
# Initialize parent classes inthe following order:
ConstrainedExpectedImprovement.__init__(self,
model=model_list,
best_f=0.0,
objective_index=0,
constraints=constraints,
maximize=False)
AcquisitionBaseToolsConstrained.__init__(
self,
model_list=model_list,
Nrestarts_eta_c=options.optimization.Nrestarts)
logger.info("Starting EIC ...")
self.dim = model_list.models[idxm['cons']].dim
self.Nrestarts = options.optimization.Nrestarts
self.algo_name = options.optimization.algo_name
self.constrained_opt = OptimizationNonLinear(
dim=self.dim,
fun_obj=self.forward,
algo_str=self.algo_name,
bounds=[[0.0] * self.dim, [1.0] * self.dim],
minimize=False,
what2optimize_str="EIC acquisition")
# self.use_nlopt = False
self.disp_info_scipy_opti = options.optimization.disp_info_scipy_opti
self._rho_conserv = options.prob_satisfaction
self.x_next, self.alpha_next = None, None
self.only_prob = False
def test_constrained_expected_improvement(self):
for dtype in (torch.float, torch.double):
# one constraint
mean = torch.tensor([[-0.5, 0.0]], device=self.device,
dtype=dtype).unsqueeze(dim=-2)
variance = torch.ones(1, 2, device=self.device,
dtype=dtype).unsqueeze(dim=-2)
mm = MockModel(MockPosterior(mean=mean, variance=variance))
module = ConstrainedExpectedImprovement(model=mm,
best_f=0.0,
objective_index=0,
constraints={1: [None, 0]})
X = torch.empty(1, 1, device=self.device, dtype=dtype) # dummy
ei = module(X)
ei_expected_unconstrained = torch.tensor(0.19780,
device=self.device,
dtype=dtype)
ei_expected = ei_expected_unconstrained * 0.5
self.assertTrue(torch.allclose(ei, ei_expected, atol=1e-4))
# check that error raised if no constraints
with self.assertRaises(ValueError):
module = ConstrainedExpectedImprovement(model=mm,
best_f=0.0,
objective_index=0,
constraints={})
# check that error raised if objective is a constraint
with self.assertRaises(ValueError):
module = ConstrainedExpectedImprovement(
model=mm,
best_f=0.0,
objective_index=0,
constraints={0: [None, 0]})
# check that error raised if constraint lower > upper
with self.assertRaises(ValueError):
module = ConstrainedExpectedImprovement(
model=mm,
best_f=0.0,
objective_index=0,
constraints={0: [1, 0]})
# three constraints
N = torch.distributions.Normal(loc=0.0, scale=1.0)
a = N.icdf(
torch.tensor(0.75)) # get a so that P(-a <= N <= a) = 0.5
mean = torch.tensor([[-0.5, 0.0, 5.0, 0.0]],
device=self.device,
dtype=dtype).unsqueeze(dim=-2)
variance = torch.ones(1, 4, device=self.device,
dtype=dtype).unsqueeze(dim=-2)
mm = MockModel(MockPosterior(mean=mean, variance=variance))
module = ConstrainedExpectedImprovement(
model=mm,
best_f=0.0,
objective_index=0,
constraints={
1: [None, 0],
2: [5.0, None],
3: [-a, a]
},
)
X = torch.empty(1, 1, device=self.device, dtype=dtype) # dummy
ei = module(X)
ei_expected_unconstrained = torch.tensor(0.19780,
device=self.device,
dtype=dtype)
ei_expected = ei_expected_unconstrained * 0.5 * 0.5 * 0.5
self.assertTrue(torch.allclose(ei, ei_expected, atol=1e-4))
# test maximize
module_min = ConstrainedExpectedImprovement(
model=mm,
best_f=0.0,
objective_index=0,
constraints={1: [None, 0]},
maximize=False,
)
ei_min = module_min(X)
ei_expected_unconstrained_min = torch.tensor(0.6978,
device=self.device,
dtype=dtype)
ei_expected_min = ei_expected_unconstrained_min * 0.5
self.assertTrue(torch.allclose(ei_min, ei_expected_min, atol=1e-4))
# test invalid onstraints
with self.assertRaises(ValueError):
ConstrainedExpectedImprovement(
model=mm,
best_f=0.0,
objective_index=0,
constraints={1: [1.0, -1.0]},
)
