priors.Wavelet

Contents

priors.Wavelet#

class bartorch.priors.Wavelet#

Bases: Regularizer

\(\ell_1\) norm of the wavelet coefficients (BART’s W).

\(\lambda \| \Psi x \|_1\). The wavelet transform \(\Psi\) is inside the proximal operator, so \(G\) is the identity.

Parameters:

  • axes (int or tuple of int) – Axes to transform, as indices into the image’s shape.

  • weight (float) – \(\lambda\).

  • joint_axes (int or tuple of int, default=()) – Axes along which a coefficient is kept or zeroed together.

  • family ({'haar', 'dau2', 'cdf44'}, default='dau2') – Wavelet family.

  • randshift (bool, default=True) – Cycle-spin the transform by a random shift at each application.

Examples

>>> Wavelet(axes=(-1, -2), weight=0.005)

Also has the methods and properties of Regularizer.

Contents