DiscreteInput requires bounds Error

[behradesg]

Dear All,

I have been trying out one of the examples of DOE using DiscreteInput and I am encountering the following error:
"ValueError: bounds must specify both lower and upper limits for each decision variable."
Here is my code:
import numpy as np
import matplotlib.pyplot as plt
from matplotlib.ticker import FormatStrFormatter

from bofire.data_models.constraints.api import (
NonlinearEqualityConstraint,
NonlinearInequalityConstraint,
LinearEqualityConstraint,
LinearInequalityConstraint,
InterpointEqualityConstraint,
)
from bofire.data_models.domain.api import Domain
from bofire.data_models.features.api import ContinuousInput, ContinuousOutput, DiscreteInput
from bofire.strategies.doe.design import find_local_max_ipopt

domain = Domain(
inputs = [
DiscreteInput(key="x1", values=[i for i in range(400, 850, 50)]),
DiscreteInput(key="x2", values=[0, 0.025, 0.05, 0.075, 0.1]),
DiscreteInput(key="x3", values=[1, 2, 3])
],
outputs = [ContinuousOutput(key="y")]
)
I tried to specify the bounds as well, but I encounter another error:
"ValidationError: 1 validation error for DiscreteInput
bounds
Extra inputs are not permitted [type=extra_forbidden, input_value=(400, 850), input_type=tuple]
For further information visit https://errors.pydantic.dev/2.8/v/extra_forbidden"

I would highly appreciate your help in this regard. Looking forward to hearing from you.

Thanks,
Behrad

[Osburg]

Hi Behrad :),

Unfortunately I cannot reproduce your first error from the code you posted. From your code I assumed that you are referring to the first example in this notebook.

When I run the code from this example replacing the original domain definition with your discrete input example the error i get is a different one
ValueError: lb and ub must either be None or have length n.

This happens since find_local_max_ipopt() only supports continuous inputs.
You can check out find_local_max_ipopt_BaB() and find_local_max_ipopt_exhaustive() if you want to use categorical inputs.

Cheers
Aaron

[Osburg]

Sure, I can try to help you. Are you interested in a list of all possible experimental combinations or in a (d-)optimal experimental design with a given number of experiments n_experiment? In the latter case: what is n_experiments in your situation?

[behradesg]

Dear Aaron,

Thank you so much. I am only interested in a list of all possible experimental combinations.

[Osburg]

Dear Behrad,

I think you get what you want to have with the following lines of code:

import itertools

x1 = [400, 450, 500, 550, 600, 650, 700, 750, 800]
x2 = [0, 0.025, 0.05, 0.075, 0.1]
x3 = [1, 2, 3]

all_combinations = list(itertools.product(x1, x2, x3))

Does this solve your problem?

[behradesg]

Thank you so much. This is what I was looking for.

[Osburg]

Sure, no problem :)