Example for using dependency constraints in discrete searchspaces¶
This example shows how a dependency constraint can be created for a discrete searchspace. For instance, some parameters might only be relevant when another parameter has a certain value. All dependencies have to be declared in a single constraint.
This example assumes some basic familiarity with using BayBE.
We thus refer to campaign for a basic example.
Necessary imports for this example¶
import os
import numpy as np
from baybe import Campaign
from baybe.constraints import DiscreteDependenciesConstraint, SubSelectionCondition
from baybe.objectives import SingleTargetObjective
from baybe.parameters import (
CategoricalParameter,
NumericalDiscreteParameter,
SubstanceParameter,
)
from baybe.searchspace import SearchSpace
from baybe.targets import NumericalTarget
from baybe.utils.dataframe import add_fake_measurements
Experiment setup¶
SMOKE_TEST = "SMOKE_TEST" in os.environ
RESOLUTION = 3 if SMOKE_TEST else 7
dict_solvent = {
"water": "O",
"C1": "C",
}
solvent = SubstanceParameter(name="Solv", data=dict_solvent, encoding="MORDRED")
switch1 = CategoricalParameter(name="Switch1", values=["on", "off"])
switch2 = CategoricalParameter(name="Switch2", values=["left", "right"])
fraction1 = NumericalDiscreteParameter(
name="Frac1", values=list(np.linspace(0, 100, RESOLUTION)), tolerance=0.2
)
frame1 = CategoricalParameter(name="FrameA", values=["A", "B"])
frame2 = CategoricalParameter(name="FrameB", values=["A", "B"])
parameters = [solvent, switch1, switch2, fraction1, frame1, frame2]
Creating the constraints¶
The constraints are handled when creating the searchspace object. It is thus necessary to define it before the searchspace creation. Note that multiple dependencies have to be included in a single constraint object.
constraint = DiscreteDependenciesConstraint(
parameters=["Switch1", "Switch2"],
conditions=[
SubSelectionCondition(selection=["on"]),
SubSelectionCondition(selection=["right"]),
],
affected_parameters=[["Solv", "Frac1"], ["FrameA", "FrameB"]],
)
Creating the searchspace and the objective¶
searchspace = SearchSpace.from_product(parameters=parameters, constraints=[constraint])
objective = SingleTargetObjective(target=NumericalTarget(name="Target_1", mode="MAX"))
Creating and printing the campaign¶
campaign = Campaign(searchspace=searchspace, objective=objective)
print(campaign)
Campaign
Meta Data
Batches done: 0
Fits done: 0
Discrete Subspace Meta Data
Recommended: 0/35
Measured: 0/35
Excluded: 0/35
SearchSpace
Search Space Type: DISCRETE
SubspaceDiscrete
Discrete Parameters
Name Type Num_Values
Encoding 0 Frac1 NumericalDiscreteParameter 3 None 1 FrameA CategoricalParameter 2 CategoricalEncoding.OHE 2 FrameB CategoricalParameter 2 CategoricalEncoding.OHE 3 Solv SubstanceParameter 2 SubstanceEncoding.MORDRED 4 Switch1 CategoricalParameter 2 CategoricalEncoding.OHE 5 Switch2 CategoricalParameter 2 CategoricalEncoding.OHE Experimental Representation Solv Switch1 … FrameA FrameB 0 C1 off … A A 1 C1 off … A A 2 C1 off … A B .. … … … … … 32 water on … A B 33 water on … B A 34 water on … B B
[35 rows x 6 columns]
Constraints
Type Affected_Paramet
0 DiscreteDepe... [Switch1, Sw...
Computational Representation
Frac1 FrameA_A ... Switch2_left Switch2_right
0 0.0 1.0 ... 1.0 0.0
1 0.0 1.0 ... 0.0 1.0
2 0.0 1.0 ... 0.0 1.0
.. ... ... ... ... ...
32 100.0 1.0 ... 0.0 1.0
33 100.0 0.0 ... 0.0 1.0
34 100.0 0.0 ... 0.0 1.0
[35 rows x 10 columns]
Objective
Type: SingleTargetObjective
Targets
Type Name ... Upper_Bound Transformation
0 NumericalTarget Target_1 ... inf None
[1 rows x 6 columns]
TwoPhaseMetaRecommender
Initial recommender
RandomRecommender
Compatibility: SearchSpaceType.HYBRID
Recommender
BotorchRecommender
Surrogate
GaussianProcessSurrogate
Supports Transfer Learning: True
Kernel factory: DefaultKernelFactory()
Acquisition function: qLogExpectedImprovement()
Compatibility: SearchSpaceType.HYBRID
Sequential continuous: False
Hybrid sampler: None
Sampling percentage: 1.0
Switch after: 1
Manual verification of the constraints¶
The following loop performs some recommendations and manually verifies the given constraints.
N_ITERATIONS = 2 if SMOKE_TEST else 5
for kIter in range(N_ITERATIONS):
print(f"\n#### ITERATION {kIter+1} ####")
print("## ASSERTS ##")
print(
f"Number entries with both switches on "
f"(expected {RESOLUTION*len(dict_solvent)*2*2}): ",
(
(campaign.searchspace.discrete.exp_rep["Switch1"] == "on")
& (campaign.searchspace.discrete.exp_rep["Switch2"] == "right")
).sum(),
)
print(
f"Number entries with Switch1 off " f"(expected {2*2}): ",
(
(campaign.searchspace.discrete.exp_rep["Switch1"] == "off")
& (campaign.searchspace.discrete.exp_rep["Switch2"] == "right")
).sum(),
)
print(
f"Number entries with Switch2 off "
f"(expected {RESOLUTION*len(dict_solvent)}):"
f" ",
(
(campaign.searchspace.discrete.exp_rep["Switch1"] == "on")
& (campaign.searchspace.discrete.exp_rep["Switch2"] == "left")
).sum(),
)
print(
"Number entries with both switches off (expected 1): ",
(
(campaign.searchspace.discrete.exp_rep["Switch1"] == "off")
& (campaign.searchspace.discrete.exp_rep["Switch2"] == "left")
).sum(),
)
rec = campaign.recommend(batch_size=5)
add_fake_measurements(rec, campaign.targets)
campaign.add_measurements(rec)
#### ITERATION 1 ####
## ASSERTS ##
Number entries with both switches on (expected 24): 24
Number entries with Switch1 off (expected 4): 4
Number entries with Switch2 off (expected 6): 6
Number entries with both switches off (expected 1): 1
#### ITERATION 2 ####
## ASSERTS ##
Number entries with both switches on (expected 24): 24
Number entries with Switch1 off (expected 4): 4
Number entries with Switch2 off (expected 6): 6
Number entries with both switches off (expected 1): 1