107 lines
6.4 KiB
Python
107 lines
6.4 KiB
Python
import torch
|
|
from typing import Union, List
|
|
|
|
def constant(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
return torch.ones_like(t_span)
|
|
|
|
def linear(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**1) * t_span**1
|
|
|
|
def quadratic(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**2) * t_span**2
|
|
|
|
def cubic(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**3) * t_span**3
|
|
|
|
def square_root(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**(1/2)) * t_span**(1/2)
|
|
|
|
def cubic_root(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**(1/3)) * t_span**(1/3)
|
|
|
|
def inverse(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return (((1/min_val)**(1/1) - 1) * 1/t_max * t_span + 1)**-1
|
|
|
|
def inverse_squared(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return (((1/min_val)**(1/2) - 1) * 1/t_max * t_span + 1)**-2
|
|
|
|
def inverse_cubed(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return (((1/min_val)**(1/3) - 1) * 1/t_max * t_span + 1)**-3
|
|
|
|
def linear_mirrored(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**1) * (-t_span + t_max)**1
|
|
|
|
def quadratic_mirrored(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**2) * (-t_span + t_max)**2
|
|
|
|
def cubic_mirrored(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**3) * (-t_span + t_max)**3
|
|
|
|
def square_root_mirrored(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**(1/2)) * (-t_span + t_max)**(1/2)
|
|
|
|
def cubic_root_mirrored(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return min_val + ((1 - min_val) / (t_max)**(1/3)) * (-t_span + t_max)**(1/3)
|
|
|
|
def inverse_mirrored(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return (((1/min_val)**(1/1) - 1) * 1/t_max * (-t_span + t_max) + 1)**-1
|
|
|
|
def inverse_squared_mirrored(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return (((1/min_val)**(1/2) - 1) * 1/t_max * (-t_span + t_max) + 1)**-2
|
|
|
|
def inverse_cubed_mirrored(t_span: Union[torch.Tensor, List[float]], t_max: float = None, min_val: float = 0.01) -> torch.Tensor:
|
|
t_span = t_span.clone().detach() if isinstance(t_span, torch.Tensor) else torch.tensor(t_span)
|
|
t_max = t_max if t_max is not None else t_span[-1]
|
|
return (((1/min_val)**(1/3) - 1) * 1/t_max * (-t_span + t_max) + 1)**-3
|
|
|
|
# Dictionary to store function references
|
|
weight_functions = {
|
|
'constant': constant,
|
|
'linear': linear,
|
|
'quadratic': quadratic,
|
|
'cubic': cubic,
|
|
'square_root': square_root,
|
|
'cubic_root': cubic_root,
|
|
'inverse': inverse,
|
|
'inverse_squared': inverse_squared,
|
|
'inverse_cubed': inverse_cubed,
|
|
'linear_mirrored': linear_mirrored,
|
|
'quadratic_mirrored': quadratic_mirrored,
|
|
'cubic_mirrored': cubic_mirrored,
|
|
'square_root_mirrored': square_root_mirrored,
|
|
'cubic_root_mirrored': cubic_root_mirrored,
|
|
'inverse_mirrored': inverse_mirrored,
|
|
'inverse_squared_mirrored': inverse_squared_mirrored,
|
|
'inverse_cubed_mirrored': inverse_cubed_mirrored,
|
|
} |