alibi_detect.utils.pytorch.kernels

DeepKernel

Inherits from: Module

Computes similarities as k(x,y) = (1-eps)k_a(proj(x), proj(y)) + epsk_b(x,y).

A forward pass takes a batch of instances x [Nx, features] and y [Ny, features] and returns the kernel matrix [Nx, Ny].

Constructor

DeepKernel(self, proj: torch.nn.modules.module.Module, kernel_a: Union[torch.nn.modules.module.Module, str] = 'rbf', kernel_b: Union[torch.nn.modules.module.Module, str, NoneType] = 'rbf', eps: Union[float, str] = 'trainable') -> None

Name	Type	Default	Description
proj	torch.nn.modules.module.Module
kernel_a	Union[torch.nn.modules.module.Module, str]	'rbf'
kernel_b	Union[torch.nn.modules.module.Module, str, None]	'rbf'
eps	Union[float, str]	'trainable'

Properties

Property	Type	Description
eps	torch.Tensor

Methods

forward

forward(x: torch.Tensor, y: torch.Tensor) -> torch.Tensor

Name	Type	Default	Description
x	torch.Tensor
y	torch.Tensor

Returns

• Type: torch.Tensor

from_config

from_config(config)

Name	Type	Default	Description
config

get_config

get_config() -> dict

Returns

• Type: dict

GaussianRBF

Inherits from: Module

Constructor

GaussianRBF(self, sigma: Optional[torch.Tensor] = None, init_sigma_fn: Optional[Callable] = None, trainable: bool = False) -> None

Name	Type	Default	Description
sigma	Optional[torch.Tensor]	None	Bandwidth used for the kernel. Needn't be specified if being inferred or trained. Can pass multiple values to eval kernel with and then average.
init_sigma_fn	Optional[Callable]	None	Function used to compute the bandwidth sigma. Used when sigma is to be inferred. The function's signature should match :py:func:~alibi_detect.utils.pytorch.kernels.sigma_median, meaning that it should take in the tensors x, y and dist and return sigma. If None, it is set to :func:~alibi_detect.utils.pytorch.kernels.sigma_median.
trainable	bool	False	Whether or not to track gradients w.r.t. sigma to allow it to be trained.

Properties

Property	Type	Description
sigma	torch.Tensor

Methods

forward

forward(x: Union[numpy.ndarray, torch.Tensor], y: Union[numpy.ndarray, torch.Tensor], infer_sigma: bool = False) -> torch.Tensor

Name	Type	Default	Description
x	Union[numpy.ndarray, torch.Tensor]
y	Union[numpy.ndarray, torch.Tensor]
infer_sigma	bool	False

Returns

• Type: torch.Tensor

from_config

from_config(config)

Instantiates a kernel from a config dictionary.

Name	Type	Default	Description
config			A kernel config dictionary.

get_config

get_config() -> dict

Returns a serializable config dict (excluding the input_sigma_fn, which is serialized in alibi_detect.saving).

Returns

• Type: dict

Functions

sigma_median

sigma_median(x: torch.Tensor, y: torch.Tensor, dist: torch.Tensor) -> torch.Tensor

Bandwidth estimation using the median heuristic :cite:t:Gretton2012.

Name	Type	Default	Description
x	torch.Tensor		Tensor of instances with dimension [Nx, features].
y	torch.Tensor		Tensor of instances with dimension [Ny, features].
dist	torch.Tensor		Tensor with dimensions [Nx, Ny], containing the pairwise distances between x and y.

Returns

• Type: torch.Tensor