autoencoder.spherical.convolution#

Module Contents#

Classes#

QuadraticNonLinearity

Special non-linearity function for data in spectral domain.

S2Convolution

Convolution between spherical signals and kernels in spectral domain.

SO3Convolution

Convolution between SO(3) signals and kernels in spectral domain.
class autoencoder.spherical.convolution.QuadraticNonLinearity(l_in, l_out, symmetric=True)#

Bases: torch.nn.Module

Special non-linearity function for data in spectral domain. Code adapated from: https://gitlab.inria.fr/ssedlar/fourier_s2cnn/-/blob/master/fourier_s2cnn/estimators/fourier_s2cnn_utils.py#L180

Parameters

  • l_in (int) – input Spherical Harmonic degree.

  • l_out (int) – output Spherical Harmonic degree.

  • symmetric (bool) – only use Spherical Harmonics of even degree. Defaults to True.

forward(self, x)#
Parameters

x (Dict[int, torch.Tensor]) –

Return type

Dict[int, torch.Tensor]

class autoencoder.spherical.convolution.S2Convolution(ti_n, te_n, l_in, b_in, b_out, symmetric=True)#

Bases: torch.nn.Module

Convolution between spherical signals and kernels in spectral domain.

Parameters

  • ti_n (int) – number of TI values

  • te_n (int) – number of TE values

  • l_in (int) – Spherical Harmonics degree

  • b_in (int) – number of input b-values

  • b_out (int) – number of output b-values

  • symmetric (bool) – only use Spherical Harmonics of even degree. Defaults to True.

forward(self, x)#
Parameters

x (Dict[int, torch.Tensor]) –

Return type

Dict[int, torch.Tensor]

class autoencoder.spherical.convolution.SO3Convolution(ti_n, te_n, l_in, b_in, b_out, symmetric=True)#

Bases: torch.nn.Module

Convolution between SO(3) signals and kernels in spectral domain.

Parameters

  • ti_n – number of TI values

  • te_n – number of TE values

  • l_in – Spherical Harmonics degree

  • b_in – number of input b-values

  • b_out – number of output b-values

  • symmetric (bool) – only use Spherical Harmonics of even degree. Defaults to True.

forward(self, x)#
Parameters

x (Dict[int, torch.Tensor]) –

Return type

Dict[int, torch.Tensor]