Moritz Wolter

I am working on a Pytorch wavelet toolbox, with the aim of full pywt compatibility here: https://github.com/v0lta/PyTorch-Wavelet-Toolbox . Pytorch supports efficient GPU processing, so you should be able to use...

Yes, we currently do not support sparse matrix transforms in 3D. I agree it would be very nice to have.

This is going to be a big project. The last time I checked, PyTorch did not provide a sparse QR solver. We already have a gram-Schmitt alternative implemented, but it's...

As always, contributions are welcome.

Dear @david-andrew, You are right, there is still a lot of NumPy code in the cwt module. We have a test to check the GPU functionality, but backprop into the...

@david-andrew I would expect the development to turn out a lot like you are describing it. Regarding the development workflow, the idea is to follow the best practices as established...

Regarding the linting, I have hardcoded version four in https://github.com/v0lta/PyTorch-Wavelet-Toolbox/commit/cc13922e090fbd76f0150ebd718bcb57d77fbb54 . I suggest we adopt the newest version when people have sorted out the problem described in https://github.com/tholo/pytest-flake8/issues/87 .

Dear @david-andrew , I have a first working prototype. See https://github.com/v0lta/PyTorch-Wavelet-Toolbox/blob/diff_cwt/examples/continuous_signal_analysis/adaptive_cwt.py , the cwt now allows backdrop into wavelets. I am still looking for a reasonable cost function for the...

I think we have no established cost function yet. The cost won't be part of the toolbox to avoid experimental features. The new cwt function from the `diff_cwt` branch supports...

Until the next release is done the new cwt is installable via the command below. ```bash pip install git+ssh://[email protected]/v0lta/PyTorch-Wavelet-Toolbox.git@diff_cwt ```