gluonts.mx.distribution.iresnet module#

class gluonts.mx.distribution.iresnet.InvertibleResnetHybridBlock(event_shape, hidden_units: int = 16, num_hidden_layers: int = 1, num_inv_iters: int = 10, ignore_logdet: bool = False, activation: str = 'lipswish', num_power_iter: int = 1, flatten: bool = False, coeff: float = 0.9, use_caching: bool = True, *args, **kwargs)[source]#

Bases: BijectionHybridBlock

Based on [BJC19], apart from f and f_inv that are swapped.

property event_dim: int#
property event_shape#
f(x: Union[NDArray, Symbol]) → Union[NDArray, Symbol][source]#

Forward transformation of iResnet.

Parameters:

x – observations

Returns:

transformed tensor ` ext{iResnet}(x)`

Return type:

Tensor

f_inv(y: Union[NDArray, Symbol]) → Union[NDArray, Symbol][source]#

Inverse transformation of iResnet

Parameters:

y – input tensor

Returns:

transformed tensor ` ext{iResnet}^{-1}(y)`

Return type:

Tensor

log_abs_det_jac(x: Union[NDArray, Symbol], y: Union[NDArray, Symbol]) → Union[NDArray, Symbol][source]#

Logarithm of the absolute value of the Jacobian determinant corresponding to the iResnet Transform.

Parameters:

  • x – input of the forward transformation or output of the inverse transform

  • y – output of the forward transform or input of the inverse transform

Returns:

Jacobian evaluated for x as input or y as output

Return type:

Tensor

gluonts.mx.distribution.iresnet.iresnet(num_blocks: int, **block_kwargs) → ComposedBijectionHybridBlock[source]#
Parameters:

  • num_blocks – number of iResnet blocks

  • block_kwargs – keyword arguments given to initialize each block object

gluonts.mx.distribution.iresnet.log_abs_det(A: Union[NDArray, Symbol]) → Union[NDArray, Symbol][source]#

Logarithm of the absolute value of matrix A :param A: Tensor matrix from which to compute the log absolute value of its

determinant

Return type:

Tensor