def f(self, x: np.ndarray) -> float:
x1, x2, t = x
if self.rate is not None:
weight = 1.0 + np.exp(-not_none(self.rate) * t)
else:
weight = 1.0
return checked_cast(float, branin(x1=x1, x2=x2)) * weight
def get_branin_data(
trial_indices: Optional[Iterable[int]] = None,
trials: Optional[Iterable[Trial]] = None,
) -> Data:
if trial_indices and trials:
raise ValueError("Expected `trial_indices` or `trials`, not both.")
if trials:
df_dicts = [
{
"trial_index":
trial.index,
"metric_name":
"branin",
"arm_name":
not_none(checked_cast(Trial, trial).arm).name,
"mean":
branin(
float(not_none(
trial.arm).parameters["x1"]), # pyre-ignore[6]
float(not_none(
trial.arm).parameters["x2"]), # pyre-ignore[6]
),
"sem":
0.0,
} for trial in trials
]
else:
df_dicts = [{
"trial_index": trial_index,
"metric_name": "branin",
"arm_name": f"{trial_index}_0",
"mean": 5.0,
"sem": 0.0,
} for trial_index in (trial_indices or [0])]
return Data(df=pd.DataFrame.from_records(df_dicts))
def testEncodeDecodeSimpleBenchmarkProblem(self):
branin_problem = get_branin_simple_benchmark_problem()
sum_problem = get_sum_simple_benchmark_problem()
new_branin_problem = object_from_json(object_to_json(branin_problem))
new_sum_problem = object_from_json(object_to_json(sum_problem))
self.assertEqual(branin_problem.f(1, 2), new_branin_problem.f(1, 2),
branin(1, 2))
self.assertEqual(sum_problem.f([1, 2]), new_sum_problem.f([1, 2]), 3)
def branin_centered(X):
# Evaluate branin problem, scaled to X \in [-1, 1]^2
# Map from [-1, 1]^2 to [[-5, 10], [0, 15]]
assert X.min() >= -1
assert X.max() <= 1
Xu = (X + 1) / 2.
Xu *= 15
Xu[:, 0] -= 5
return branin(Xu)
def f(self, x: np.ndarray) -> Mapping[str, Any]:
if self.index < len(FIDELITY):
self.index += 1
x1, x2 = x
fidelity = FIDELITY[self.index]
fidelity_penalty = random() * math.pow(1.0 - fidelity, 2.0)
mean = checked_cast(float, branin(x1=x1, x2=x2)) - fidelity_penalty
return {"mean": mean, "fidelity": fidelity}
def f(self, x: np.ndarray, timestamp: int) -> Mapping[str, Any]:
x1, x2 = x
if self.rate is not None:
weight = 1.0 + np.exp(-not_none(self.rate) * timestamp)
else:
weight = 1.0
mean = checked_cast(float, branin(x1=x1, x2=x2)) * weight
return {"mean": mean, "timestamp": timestamp}
def get_branin_data_batch(batch: BatchTrial) -> Data:
return Data(
pd.DataFrame({
"arm_name": [arm.name for arm in batch.arms],
"metric_name":
"branin",
"mean": [
# pyre-ignore[6]: This function can fail if a parameter value
# does not support conversion to float.
branin(float(arm.parameters["x1"]),
float(arm.parameters["x2"])) for arm in batch.arms
],
"sem":
0.1,
}))
def testEncodeDecodeSimpleBenchmarkProblem(self):
branin_problem = get_branin_simple_benchmark_problem()
sum_problem = get_sum_simple_benchmark_problem()
new_branin_problem = object_from_json(object_to_json(branin_problem))
new_sum_problem = object_from_json(object_to_json(sum_problem))
self.assertEqual(branin_problem.f(1, 2), new_branin_problem.f(1, 2),
branin(1, 2))
self.assertEqual(sum_problem.f([1, 2]), new_sum_problem.f([1, 2]), 3)
# Test using `from_botorch`.
ackley_problem = SimpleBenchmarkProblem(f=from_botorch(Ackley()),
noise_sd=0.0,
minimize=True)
new_ackley_problem = object_from_json(object_to_json(ackley_problem))
self.assertEqual(ackley_problem.f(1, 2), new_ackley_problem.f(1, 2),
ackley(1, 2))
def test_branin(self):
self.assertEqual(branin.name, "Branin")
self.assertAlmostEqual(branin(1, 2), 21.62763539206238)
self.assertAlmostEqual(branin(x1=1, x2=2), 21.62763539206238)
self.assertAlmostEqual(branin(np.array([1, 2])), 21.62763539206238)
self.assertAlmostEqual(branin(np.array([[1, 2], [1, 2]]))[0], 21.62763539206238)
self.assertAlmostEqual(branin.minimums[0][0], -np.pi)
self.assertAlmostEqual(branin.fmin, 0.397887, places=6)
self.assertAlmostEqual(branin.fmax, 308.129, places=3)
self.assertAlmostEqual(branin.fmax, branin(-5, 0), places=3)
self.assertEqual(branin.domain[0], (-5, 10))
self.assertEqual(branin.required_dimensionality, 2)
with self.assertRaisesRegex(NotImplementedError, "Branin does not specify"):
branin.maximums
with self.assertRaisesRegex(ValueError, "Synthetic function call"):
branin(np.array([[[1, 3]]]))
def testEncodeDecodeSimpleBenchmarkProblem(self):
branin_problem = get_branin_simple_benchmark_problem()
sum_problem = get_sum_simple_benchmark_problem()
new_branin_problem = object_from_json(
object_to_json(
branin_problem,
encoder_registry=DEPRECATED_ENCODER_REGISTRY,
class_encoder_registry=DEPRECATED_CLASS_ENCODER_REGISTRY,
),
decoder_registry=DEPRECATED_DECODER_REGISTRY,
class_decoder_registry=DEPRECATED_CLASS_DECODER_REGISTRY,
)
new_sum_problem = object_from_json(
object_to_json(
sum_problem,
encoder_registry=DEPRECATED_ENCODER_REGISTRY,
class_encoder_registry=DEPRECATED_CLASS_ENCODER_REGISTRY,
),
decoder_registry=DEPRECATED_DECODER_REGISTRY,
class_decoder_registry=DEPRECATED_CLASS_DECODER_REGISTRY,
)
self.assertEqual(
branin_problem.f(1, 2), new_branin_problem.f(1, 2), branin(1, 2)
)
self.assertEqual(sum_problem.f([1, 2]), new_sum_problem.f([1, 2]), 3)
# Test using `from_botorch`.
ackley_problem = SimpleBenchmarkProblem(
f=from_botorch(Ackley()), noise_sd=0.0, minimize=True
)
new_ackley_problem = object_from_json(
object_to_json(
ackley_problem,
encoder_registry=DEPRECATED_ENCODER_REGISTRY,
class_encoder_registry=DEPRECATED_CLASS_ENCODER_REGISTRY,
),
decoder_registry=DEPRECATED_DECODER_REGISTRY,
class_decoder_registry=DEPRECATED_CLASS_DECODER_REGISTRY,
)
self.assertEqual(
ackley_problem.f(1, 2), new_ackley_problem.f(1, 2), ackley(1, 2)
)
def f(self, x: np.ndarray) -> float:
x1, x2 = x
return checked_cast(float, branin(x1=x1, x2=x2))
def f(self, x: np.ndarray) -> float:
x1, x2, fidelity = x
fidelity_penalty = random() * math.pow(1.0 - fidelity, 2.0)
return checked_cast(float, branin(x1=x1, x2=x2)) - fidelity_penalty
