tfmri.convex.ConvexFunctionL1Wavelet

class ConvexFunctionL1Wavelet(domain_shape, wavelet, mode='symmetric', level=None, axes=None, scale=None, dtype=tf.float32, name=None)[source]

Bases: tensorflow_mri.python.ops.convex_ops.ConvexFunctionLinearOperatorComposition

A ConvexFunction representing an L1 wavelet regularization term.

For a given input \(x\), computes \(\lambda \left\| Dx \right\|_1\), where \(\lambda\) is a scaling factor and \(D\) is a wavelet decomposition operator (see tfmri.linalg.LinearOperatorWavelet).

Parameters

  • domain_shape – A 1D integer tf.Tensor. The domain shape of this linear operator. This operator may have multiple domain dimensions.

  • wavelet –

    A str or a `pywt.Wavelet`_, or a list thereof. When passed a list, different wavelets are applied along each axis in axes.

  • mode –

    A str. The padding or signal extension mode. Must be one of the values supported by tfmri.signal.wavedec. Defaults to 'symmetric'.

  • level – An int >= 0. The decomposition level. If None (default), the maximum useful level of decomposition will be used (see tfmri.signal.max_wavelet_level).

  • axes –

    A list of int. The axes over which the DWT is computed. Axes refer only to domain dimensions without regard for the batch dimensions. Defaults to None (all domain dimensions).

  • scale – A float. A scaling factor.

  • dtype – A tf.dtypes.DType. The dtype of the inputs.

  • name – A name for this ConvexFunction.

Initialize this ConvexFunction.