Skip to content

Commit

Permalink
Add input constructor for qHypervolumeKnowledgeGradient (#2501)
Browse files Browse the repository at this point in the history 
Summary:
Pull Request resolved: #2501

Adds new input constructors for `qHypervolumeKnowledgeGradient`.

Reviewed By: Balandat

Differential Revision: D62046832

fbshipit-source-id: acbb823c37774ec104fa86688ddbc3fed6996b8c
  • Loading branch information
@ltiao @facebook-github-bot
ltiao authored and facebook-github-bot committed Sep 12, 2024
1 parent 4d49bf7 commit 127fc08
Show file tree
Hide file tree
Showing 2 changed files with 214 additions and 9 deletions.
65 changes: 65 additions & 0 deletions botorch/acquisition/input_constructors.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -62,6 +62,10 @@
qExpectedHypervolumeImprovement,
qNoisyExpectedHypervolumeImprovement,
)
from botorch.acquisition.multi_objective.hypervolume_knowledge_gradient import (
_get_hv_value_function,
qHypervolumeKnowledgeGradient,
)
from botorch.acquisition.multi_objective.logei import (
qLogExpectedHypervolumeImprovement,
qLogNoisyExpectedHypervolumeImprovement,
Expand Down Expand Up @@ -1270,6 +1274,67 @@ def construct_inputs_qKG(
return inputs_qkg


def _get_ref_point(
objective_thresholds: Tensor,
objective: Optional[MCMultiOutputObjective] = None,
) -> Tensor:

if objective is None:
ref_point = objective_thresholds
elif isinstance(objective, RiskMeasureMCObjective):
ref_point = objective.preprocessing_function(objective_thresholds)
else:
ref_point = objective(objective_thresholds)

return ref_point


@acqf_input_constructor(qHypervolumeKnowledgeGradient)
def construct_inputs_qHVKG(
model: Model,
training_data: MaybeDict[SupervisedDataset],
bounds: list[tuple[float, float]],
objective_thresholds: Tensor,
objective: Optional[MCMultiOutputObjective] = None,
posterior_transform: Optional[PosteriorTransform] = None,
num_fantasies: int = 8,
num_pareto: int = 10,
**optimize_objective_kwargs: TOptimizeObjectiveKwargs,
) -> dict[str, Any]:
r"""Construct kwargs for `qKnowledgeGradient` constructor."""

X = _get_dataset_field(training_data, "X", first_only=True)
_bounds = torch.as_tensor(bounds, dtype=X.dtype, device=X.device)

ref_point = _get_ref_point(
objective_thresholds=objective_thresholds, objective=objective
)

acq_function = _get_hv_value_function(
model=model,
ref_point=ref_point,
use_posterior_mean=True,
objective=objective,
)

_, current_value = optimize_objective(
model=model,
bounds=_bounds.t(),
q=num_pareto,
acq_function=acq_function,
**optimize_objective_kwargs,
)

return {
"model": model,
"objective": objective,
"ref_point": ref_point,
"num_fantasies": num_fantasies,
"num_pareto": num_pareto,
"current_value": current_value.detach().cpu().max(),
}


@acqf_input_constructor(qMultiFidelityKnowledgeGradient)
def construct_inputs_qMFKG(
model: Model,
Expand Down
158 changes: 149 additions & 9 deletions test/acquisition/test_input_constructors.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -72,6 +72,10 @@
qExpectedHypervolumeImprovement,
qNoisyExpectedHypervolumeImprovement,
)
from botorch.acquisition.multi_objective.hypervolume_knowledge_gradient import (
qHypervolumeKnowledgeGradient,
)

from botorch.acquisition.multi_objective.logei import (
qLogExpectedHypervolumeImprovement,
qLogNoisyExpectedHypervolumeImprovement,
Expand Down Expand Up @@ -222,7 +226,7 @@ def test_optimize_objective(self, mock_optimize_acqf):
mock_model = self.mock_model
bounds = torch.rand(2, len(self.bounds))

with self.subTest("scalarObjective_acqusitionFunction"):
with self.subTest("scalarObjective_acquisitionFunction"):
optimize_objective(
model=mock_model,
bounds=bounds,
Expand Down Expand Up @@ -1072,8 +1076,9 @@ def test_construct_inputs_qEHVI(self) -> None:
self.assertIs(kwargs["constraints"], constraints)
self.assertEqual(kwargs["eta"], 1e-2)

with self.subTest("block designs"), self.assertRaisesRegex(
ValueError, "Field `X` must be shared"
with (
self.subTest("block designs"),
self.assertRaisesRegex(ValueError, "Field `X` must be shared"),
):
c(
model=mm,
Expand Down Expand Up @@ -1291,6 +1296,119 @@ def test_construct_inputs_kg(self) -> None:
)
self.assertNotIn("current_value", kwargs)

def test_construct_inputs_hvkg(self) -> None:
model = mock.Mock()
current_value = torch.tensor(1.23)

objective_thresholds = torch.rand(2)
objective = IdentityMCMultiOutputObjective()

get_kwargs = get_acqf_input_constructor(qHypervolumeKnowledgeGradient)

with (
mock.patch(
target="botorch.acquisition.input_constructors._get_hv_value_function",
) as mock_get_hv_value_function,
mock.patch(
target="botorch.acquisition.input_constructors.optimize_acqf",
return_value=(None, current_value),
) as mock_optimize_acqf,
):

kwargs = get_kwargs(
model=model,
training_data=self.blockX_blockY,
objective_thresholds=objective_thresholds,
objective=objective,
bounds=self.bounds,
num_fantasies=33,
num_pareto=11,
)

self.assertEqual(
mock_get_hv_value_function.call_args.kwargs["model"], model
)
self.assertEqual(
mock_get_hv_value_function.call_args.kwargs["objective"], objective
)
self.assertTrue(
torch.equal(
mock_get_hv_value_function.call_args.kwargs["ref_point"],
objective_thresholds,
)
)

# check that `optimize_acqf` is called with the desired value function
self.assertEqual(
mock_optimize_acqf.call_args.kwargs["acq_function"],
mock_get_hv_value_function(),
)

self.assertLessEqual(
{
"model",
"ref_point",
"num_fantasies",
"num_pareto",
"objective",
"current_value",
},
set(kwargs.keys()),
)
self.assertEqual(kwargs["num_fantasies"], 33)
self.assertEqual(kwargs["num_pareto"], 11)
self.assertEqual(kwargs["current_value"], current_value)
self.assertTrue(torch.equal(kwargs["ref_point"], objective_thresholds))

with self.subTest("custom objective"):
weights = torch.rand(2)
objective = WeightedMCMultiOutputObjective(weights=weights)
with mock.patch(
target="botorch.acquisition.input_constructors.optimize_acqf",
return_value=(None, current_value),
) as mock_optimize_acqf:
kwargs = get_kwargs(
model=model,
training_data=self.blockX_blockY,
objective_thresholds=objective_thresholds,
objective=objective,
bounds=self.bounds,
num_fantasies=33,
num_pareto=11,
)
self.assertIsInstance(kwargs["objective"], WeightedMCMultiOutputObjective)
self.assertTrue(
torch.equal(kwargs["ref_point"], objective_thresholds * weights)
)

with self.subTest("risk measures"):
for use_preprocessing in (True, False):
objective = MultiOutputExpectation(
n_w=3,
preprocessing_function=(
WeightedMCMultiOutputObjective(torch.tensor([-1.0, -1.0]))
if use_preprocessing
else None
),
)
with mock.patch(
target="botorch.acquisition.input_constructors.optimize_acqf",
return_value=(None, current_value),
) as mock_optimize_acqf:
kwargs = get_kwargs(
model=model,
training_data=self.blockX_blockY,
objective_thresholds=objective_thresholds,
objective=objective,
bounds=self.bounds,
num_fantasies=33,
num_pareto=11,
)
expected_obj_t = (
-objective_thresholds if use_preprocessing else objective_thresholds
)
self.assertTrue(torch.equal(kwargs["ref_point"], expected_obj_t))

def test_construct_inputs_mes(self) -> None:
func = get_acqf_input_constructor(qMaxValueEntropy)
n, d, m = 5, 2, 1
Expand Down Expand Up @@ -1429,12 +1547,17 @@ def test_construct_inputs_mfmes(self) -> None:
"cost_intercept": 0.321,
}
input_constructor = get_acqf_input_constructor(qMultiFidelityMaxValueEntropy)
with mock.patch(
target="botorch.acquisition.input_constructors.construct_inputs_mf_base",
return_value={"foo": 0},
), mock.patch(
target="botorch.acquisition.input_constructors.construct_inputs_qMES",
return_value={"bar": 1},
with (
mock.patch(
target=(
"botorch.acquisition.input_constructors.construct_inputs_mf_base"
),
return_value={"foo": 0},
),
mock.patch(
target="botorch.acquisition.input_constructors.construct_inputs_qMES",
return_value={"bar": 1},
),
):
inputs_mfmes = input_constructor(**constructor_args)
inputs_test = {"foo": 0, "bar": 1, "num_fantasies": 64}
Expand Down Expand Up @@ -1603,6 +1726,23 @@ def setUp(self, suppress_input_warnings: bool = True) -> None:
"training_data": self.blockX_blockY,
},
)

X = torch.rand(3, 2)
Y1 = torch.rand(3, 1)
Y2 = torch.rand(3, 1)
m1 = SingleTaskGP(X, Y1)
m2 = SingleTaskGP(X, Y2)
model_list = ModelListGP(m1, m2)

self.cases["HV Look-ahead"] = (
[qHypervolumeKnowledgeGradient],
{
"model": model_list,
"training_data": self.blockX_blockY,
"bounds": bounds,
"objective_thresholds": objective_thresholds,
},
)
pref_model = self.mock_model
pref_model.dim = 2
pref_model.datapoints = torch.tensor([])
Expand Down

0 comments on commit 127fc08

Please sign in to comment.