104 lines
3.4 KiB
Python
104 lines
3.4 KiB
Python
from typing import Any, Dict
|
|
import ast
|
|
import torch
|
|
from torch.distributions import StudentT
|
|
import scipy.stats as scst
|
|
|
|
class StudentTEx(StudentT):
|
|
"""
|
|
A class to extend StudentT to add missing functionality.
|
|
|
|
pragma: skip_doc
|
|
"""
|
|
def cdf(self, x: torch.Tensor):
|
|
"""
|
|
Computes the cumulative distribution function (CDF) of the StudentT distribution.
|
|
|
|
Args:
|
|
x (torch.Tensor): Input tensor containing values where the CDF is evaluated.
|
|
|
|
Returns:
|
|
torch.Tensor: Tensor containing the cumulative probabilities.
|
|
"""
|
|
a = x.detach().cpu().numpy()
|
|
df = self.df.detach().cpu().item()
|
|
loc = self.loc.detach().cpu().numpy()
|
|
scale = self.scale.detach().cpu().numpy()
|
|
outputs = scst.t.cdf((a - loc)/scale, df)
|
|
return torch.tensor(outputs, dtype=torch.float32).to(x.device)
|
|
|
|
def icdf(self, q: torch.Tensor):
|
|
"""
|
|
Computes the inverse of cumulative distribution function (CDF) of the StudentT distribution.
|
|
|
|
Args:
|
|
q (torch.Tensor): Input tensor containing values where the ICDF is evaluated.
|
|
|
|
Returns:
|
|
torch.Tensor: Tensor containing the inverse of cumulative probabilities.
|
|
"""
|
|
a = q.detach().cpu().numpy()
|
|
df = self.df.detach().cpu().item()
|
|
loc = self.loc.detach().cpu().numpy()
|
|
scale = self.scale.detach().cpu().numpy()
|
|
outputs = scst.t.ppf(a, df) * scale + loc
|
|
return torch.tensor(outputs, dtype=torch.float32).to(q.device)
|
|
|
|
|
|
class PtdStudentT:
|
|
"""
|
|
Instantiates a StudentT distribution object.
|
|
|
|
Args:
|
|
df (str): Degree of freedom. You can specify a scalar or a tuple of float.
|
|
loc (str): Rate. Mean of the distsribution. You can specify a scalar or a tuple of float.
|
|
scale (str): Rate. Scale of the distribution. You can specify a scalar or a tuple of float.
|
|
|
|
category: PyTorch wrapper - Distribution
|
|
"""
|
|
|
|
@classmethod
|
|
def INPUT_TYPES(cls) -> Dict[str, Any]:
|
|
"""
|
|
Defines the input types for the function.
|
|
|
|
Returns:
|
|
Dict[str, Any]: A dictionary specifying the required input types.
|
|
"""
|
|
return {
|
|
"required": {
|
|
"df": ("STRING", {"default": ""}),
|
|
"loc": ("STRING", {"default": ""}),
|
|
"scale": ("STRING", {"default": ""})
|
|
}
|
|
}
|
|
|
|
RETURN_TYPES: tuple = ("PTDISTRIBUTION",)
|
|
FUNCTION: str = "f"
|
|
CATEGORY: str = "Distribution"
|
|
|
|
def f(self,
|
|
df: str,
|
|
loc: str,
|
|
scale: str) -> tuple:
|
|
"""
|
|
Instantiates a StudentT distribution object.
|
|
|
|
Args:
|
|
df (str): Degree of freedom. You can specify a scalar or a tuple of float.
|
|
loc (str): Rate. Mean of the distsribution. You can specify a scalar or a tuple of float.
|
|
scale (str): Rate. Standard deviation of the distribution. You can specify a scalar or a tuple of float.
|
|
|
|
Returns:
|
|
tuple: (torch.distributions.studentT.StudentT)
|
|
"""
|
|
d = ast.literal_eval(df)
|
|
mu = ast.literal_eval(loc)
|
|
s = ast.literal_eval(scale)
|
|
|
|
dist = StudentTEx(
|
|
df=torch.tensor(d, dtype=torch.float32),
|
|
loc=torch.tensor(mu, dtype=torch.float32),
|
|
scale=torch.tensor(s, dtype=torch.float32))
|
|
return (dist,)
|