Descriptors¶

Descriptors specify what kinds of values are valid for a parameter. There are currently, 2 variants: RangeDescriptor and CategoryDescriptor.

(Note that even though their names end with Parameter, they are actually descriptors, not parameters.)

from besos.parameters import (
    RangeDescriptor,
    CategoryDescriptor,
    Parameter,
)
from besos.problem import Problem
from besos import sampling
from besos.evaluator import EvaluatorGeneric

RangeDescriptors¶

\(min \leq x \leq max\)

zero_to_one_exclusive = RangeDescriptor(min_val=0.01, max_val=0.99)

CategoryDescriptors¶

A list of options.

text_example = CategoryDescriptor(options=["a", "b", "c", "other"])
single_digit_integers = CategoryDescriptor(options=range(10))

Sampling¶

These descriptors can be used to make Parameters. Then we can generate samples.

parameters = [
    Parameter(value_descriptors=zero_to_one_exclusive, name="0-1"),
    Parameter(value_descriptors=single_digit_integers, name="single digit"),
    Parameter(value_descriptors=text_example, name="text"),
]
problem = Problem(parameters, outputs=["output"])

samples = sampling.dist_sampler(sampling.lhs, problem, num_samples=10)
samples

	0-1	single digit	text
0	0.380019	9	a
1	0.586379	3	a
2	0.204224	7	b
3	0.616832	5	c
4	0.424950	1	c
5	0.029658	2	c
6	0.796134	8	other
7	0.913143	6	other
8	0.290977	4	b
9	0.758265	0	a

Evaluation¶

Since we did not specify selectors for the parameters, we cannot evaluate them using an EnergyPlus simulation. Instead, we will use a custom evaluation function.

def evaluation_function(values):
    x, y, z = values
    if z == "other":
        return (x,)
    else:
        return (x * y,)


evaluator = EvaluatorGeneric(evaluation_function, problem)
# The evaluator will use this objective by default
outputs = evaluator.df_apply(samples, keep_input=True)
# outputs is a pandas dataframe with one column since only one objective was requested

HBox(children=(FloatProgress(value=0.0, description='Executing', max=10.0, style=ProgressStyle(description_wid…

outputs

	0-1	single digit	text	output
0	0.380019	9	a	3.420167
1	0.586379	3	a	1.759137
2	0.204224	7	b	1.429571
3	0.616832	5	c	3.084161
4	0.424950	1	c	0.424950
5	0.029658	2	c	0.059316
6	0.796134	8	other	0.796134
7	0.913143	6	other	0.913143
8	0.290977	4	b	1.163907
9	0.758265	0	a	0.000000