v0.1.5

[v0lta]

What's new:

• added additional test cases for our wavedec3 and 1D-matrix transformation code.
  • 3d matrix transformation support. The new matrix-wavedecs3 returns a list of dictionaries like https://pywavelets.readthedocs.io/en/latest/ref/nd-dwt-and-idwt.html#multilevel-decomposition-wavedecn .
  • The separable_conv_transform module now provides support for separable padded transforms.
• improved handling of the internal "channel"-dimension for the padded 2d-fwt case. Output tensors from wavedec2 now have the shape [batch, height, width] instead of [batch, 1, height, width].
• Added shape checks for the transformed coefficients, for more informative error messages.
• Added tests for just in time compilation. See test/test_jit.py for more details.

Breaking changes:

• since https://arxiv.org/pdf/2210.14874.pdf found them to work well, the faster separable transforms are now the default for all boundary wavelet computations. If you require wavedec2 and MatrixWavedec2 to produce identical coefficients in the Haar case, set separable to False.
• to simplify the documentation _get_pad is now private.
• To ensure our API is compatible with pywt the padded analysis transforms argument order has changed. The order is now always data, wavelet, mode, level.

[v0lta]

https://github.com/v0lta/PyTorch-Wavelet-Toolbox/pull/50 has the separable conv transforms for v.0.1.5 (work in progress).

[v0lta]

Thanks, Felix. Not having a publicly available function might be better since we do not support n-dimensional inputs anywhere else. The toolbox feels more consistent this way.

[felixblanke]

It bugged me that the separable attribute of WaveletPacket2D was ignored for convolutional transforms, so I added support for it. However I am not really happy with the necessary conversions between the [Tensor, Tuple[Tensor, ...]] and [Tensor, Dict[str, Tensor]] formats.

Btw: I have no clue why one test failed for cc74842

[felixblanke]

Currently, the default for WaveletPacket2D are non-separable transforms. Should we switch that to separable as well or keep it at non-separable?

[v0lta]

I think the default should ensure results are consistent with the pywt-packets. Looking at https://github.com/v0lta/PyTorch-Wavelet-Toolbox/blob/d4263467be00b64ae3074681e6593550a42d7540/src/ptwt/packets.py#L257 tells me we can keep the reflect default and have separable equal True. I think that makes lots of sense. Great Idea, @felixblanke !

[v0lta]

Ah, I see you added support for separable conv transforms for the 2D packets. Thank you! In this case, I would leave the defaults unchanged because I expect most users to want pywt-compatible results.

[v0lta]

The jit tests run just fine with torch 2.0 . I have tested this. https://github.com/v0lta/PyTorch-Wavelet-Toolbox/issues/51 is now closed. We should merge. Other changes should go into future releases.