Add unscaled option for embeddings (at least ASE/LSE)

Is your feature request related to a problem? Please describe.

Working on something now where I often want the unscaled version of the ASE/LSE embedding, that is, just the left or right singular vectors without scaling each by sqrt(singular value).

Describe the solution you’d like

Would like a scaled=False keyword argument, default kept as it is.

Describe alternatives you’ve considered

Obviously easy to unscale after getting the current embedding, just less convenient. Does make the embeddings even closer to just SVD, but ASE at least has diag_aug, and LSE has some preprocessing to form the Laplacian.

Provide references (if applicable)

At least one of the very early embedding papers had unscaled as an option: https://arxiv.org/abs/1108.2228

Additional context

Would also make code in MASE simpler I think? https://github.com/microsoft/graspologic/blob/dev/graspologic/embed/mase.py#L158 could just use the new ASE, passing down scaled

[rajpratyush]

@bdpedigo i dont care if graspologic comes in gsoc or not i will be actively contributing to it just i hoped with gsoc i will be able to showcase my work to my resume

[daxpryce]

@rajpratyush if there’s any other ways we can help you highlight your contributions for your resume, let us know. We really appreciate your contributions and work on this project with us - you’ve cleaned up so many of our “we want to get to this but don’t know when we’ll have the time” tasks that we’d be way worse off without all the help you’ve given.

So if there’s some other way we can assist, please reach out - it’s the least we can do!