"""Contains the surrogate modelling base class
The defined base class for surrogate modelling.
"""
import chaospy
import numpy as np
from ..data_model import ParameterDataType
from .calibration_base import CalibrationWorkflowBase
[docs]
class SurrogateBase(CalibrationWorkflowBase):
"""The surrogate modelling base class."""
[docs]
def sample_parameters(self, n_samples: int) -> np.ndarray:
"""Sample from the parameter space.
Args:
n_samples (int): The number of samples.
Returns:
np.ndarray: The sampled parameter values.
"""
X = self.parameters.sample(n_samples, rule="sobol").T
n_replicates = self.specification.n_replicates
if n_replicates > 1:
X = np.repeat(X, n_replicates, axis=0)
self.rng.shuffle(X)
return X
[docs]
def specify(self) -> None:
"""Specify the parameters of the model calibration procedure."""
self.names = []
self.data_types = []
parameters = []
parameter_spec = self.specification.parameter_spec.parameters
for spec in parameter_spec:
parameter_name = spec.name
data_type = spec.data_type
if data_type == ParameterDataType.CONSTANT:
parameter_value = spec.parameter_value
self.constants[parameter_name] = parameter_value
continue
elif data_type == ParameterDataType.CATEGORICAL:
lower_bound, upper_bound = self.set_categorical_parameter(spec)
parameter = chaospy.DiscreteUniform(lower_bound, upper_bound - 1)
else:
distribution_name = (
spec.distribution_name.replace("_", " ").title().replace(" ", "")
)
distribution_args = spec.distribution_args
if distribution_args is None:
distribution_args = []
distribution_kwargs = spec.distribution_kwargs
if distribution_kwargs is None:
distribution_kwargs = {}
dist_instance = getattr(chaospy, distribution_name)
parameter = dist_instance(*distribution_args, **distribution_kwargs)
self.names.append(parameter_name)
self.data_types.append(data_type)
parameters.append(parameter)
self.parameters = chaospy.J(*parameters)