Aaron Meyer

This is outside my expertise but seems reasonable. I would suggest profiling to see if there is a performance hit, since it could really matter depending on the context. For...

@bakhtos I enabled tests on here so you can find remaining syntax issues. These all look like pretty simple fixes.

(Sorry for the noise. I accidentally ran black.)

@cohenjer good point about having one call before any iterations. I'll add that. I will fill this out with documentation and an interface definition, then open an actual PR.

@cohenjer good point about having one call before any iterations. I'll add that. I will fill this out with documentation and an interface definition, then open an actual PR.

> About the question on what argument should the callback function take, in fine indeed we could use a dictionary, but maybe we can go for a simple version of...

Thanks for the input all. I completely agree with @cohenjer and @MarieRoald's comments. I wanted to add the specific capabilities that this would enable for me, and the alternative that...

@cohenjer you read my mind. I was going to post about both of these issues: - The rule I have been following is that the inputs to the callback should...

> Thanks @Lili-Zheng-stat! > > The issue in the tests comes from the fact that we changed the SVD interface in #429, the best would be to use `tensorly.tenalg.svd_interface`. @aarmey...

@cohenjer @caglayantuna I had another idea after we met about this. Perhaps rather than adding another optional return, we should add a callbacks interface (e.g., see [`scipy.optimize.minimize`](https://docs.scipy.org/doc/scipy/reference/generated/scipy.optimize.minimize.html))? This would provide...