Jean Kossaifi

Thanks for looking at it @yngvem! Currently if r > n I simply skipped the factor. I was thinking of having at least a simple heuristic, e.g. orthogonalizing the first...

Thanks @caglayantuna, merging!

We currently don't have this, would be great to add if you are interested in contributing! The closest we currently have is the `sparsity` argument in `parafac`, which allows to...

@DylanMannKrzisnik that would be great! We already have proximal operators, including soft and SVD thresholding: https://github.com/tensorly/tensorly/blob/f72be606027f136c35da7c2d3e970e20d1f2a1ef/tensorly/tenalg/proximal.py#L9 For usage examples, you can check the [robust tensor PCA](https://github.com/tensorly/tensorly/blob/master/tensorly/decomposition/robust_decomposition.py) that uses them.

Yes I agree. Let's decide on the API https://github.com/tensorly/lab/pull/1 and merge #284

[Updated link for Slack](https://join.slack.com/t/tensorly/shared_invite/zt-wqnts2sk-wbiRX6ml~Xt6~GDYWRPFfg)

Agreed, though generating random factorized tensors is involved if we want to maintain some properties on the reconstructed tensor (see e.g. in tensorly-torch, we allow to generate random tensors such...

@yngvem @MarieRoald -- I really like it! Currently I essentially have been relying on a backend system (both for the actual backend and the `tenalg` module). I guess the general...

Yes, I agree with both your points @aarmey and @yngvem - to summarize, there's the performance price of dynamic dispatch, this is something we'd need to handle as efficiently as...

@RichieHakim this is very interesting - it looks like it might be a good match for [TensorLy-Torch](https://github.com/tensorly/torch) as it's Pytorch only. Would you be interested in contributing it there?