from typing import List, Dict, Any, Union
import gpytorch
from numpy.typing import NDArray
import torch
from csle_common.dao.system_identification.gp_regression_model_with_gauissan_noise import \
GPRegressionModelWithGaussianNoise
from csle_base.json_serializable import JSONSerializable
[docs]class GPConditional(JSONSerializable):
"""
A DTO representing a Gaussian process conditional distribution
"""
def __init__(self, conditional_name: str, metric_name: str,
sample_space: List[int],
observed_x: List[Union[float, int]], observed_y: List[Union[float, int]],
scale_parameter: float, noise_parameter: float) -> None:
"""
Initializes the DTO
:param conditional_name: the name of the conditional
:param metric_name: the name of the metric
:param sample_space: the sample space (the domain of the distribution)
:param observed_x: the observed x samples
:param observed_y: the observed y samples
:param scale_parameter: the scale parameter for the scale kernel
:param noise_parameter: the noise parameter for the Gaussian noise likelihood
"""
self.conditional_name = conditional_name
self.metric_name = metric_name
self.sample_space = sample_space
self.observed_x = observed_x
self.observed_y = observed_y
self.scale_parameter = scale_parameter
self.noise_parameter = noise_parameter
self.distribution: List[NDArray[Any]] = []
[docs] @staticmethod
def from_dict(d: Dict[str, Any]) -> "GPConditional":
"""
Converts a dict representation of the DTO into an instance
:param d: the dict to convert
:return: the converted instance
"""
return GPConditional(
conditional_name=d["conditional_name"], metric_name=d["metric_name"],
sample_space=d["sample_space"], observed_x=d["observed_x"], observed_y=d["observed_y"],
scale_parameter=d["scale_parameter"], noise_parameter=d["noise_parameter"]
)
[docs] def to_dict(self) -> Dict[str, Any]:
"""
:return: a dict representation of the DTO
"""
d: Dict[str, Any] = {}
d["conditional_name"] = self.conditional_name
d["metric_name"] = self.metric_name
d["sample_space"] = self.sample_space
d["observed_x"] = self.observed_x
d["observed_y"] = self.observed_y
d["scale_parameter"] = self.scale_parameter
d["noise_parameter"] = self.noise_parameter
d["distribution"] = self.distribution
return d
def __str__(self) -> str:
"""
:return: a string representation of the DTO
"""
return f"conditional_name:{self.conditional_name}, metric_name: {self.metric_name}, " \
f"sample_space: {self.sample_space}, observed_x: {self.observed_x}, observed_y: {self.observed_y}," \
f"scale_parameter: {self.scale_parameter}, noise_parameter: {self.noise_parameter}"
[docs] @staticmethod
def from_json_file(json_file_path: str) -> "GPConditional":
"""
Reads a json file and converts it to a DTO
:param json_file_path: the json file path
:return: the converted DTO
"""
import io
import json
with io.open(json_file_path, 'r') as f:
json_str = f.read()
return GPConditional.from_dict(json.loads(json_str))
[docs] def generate_distribution(self) -> None:
"""
Updates the distribution
:return: None
"""
self.sample_space.sort()
self.distribution = list(self.generate_distributions_for_samples(samples=self.sample_space).tolist())
[docs] def generate_distributions_for_samples(self, samples) -> Any:
"""
Generates distributions for a given sample
:param samples: the sample
:return: the created distributions
"""
samples = torch.tensor(samples)
likelihood = gpytorch.likelihoods.GaussianLikelihood()
model = GPRegressionModelWithGaussianNoise(torch.tensor(self.observed_x), torch.tensor(self.observed_y),
likelihood)
model.covar_module.base_kernel.lengthscale = torch.tensor(self.scale_parameter)
model.likelihood.noise = torch.tensor(self.noise_parameter)
model.eval()
likelihood.eval()
# Make predictions by feeding model through likelihood
with torch.no_grad(), gpytorch.settings.fast_pred_var():
test_x = torch.tensor(samples)
dist = likelihood(model(test_x)).mean.numpy()
return dist