CASC

[loftusa]

tutorial:
https://github.com/microsoft/graspologic/blob/8b428688f528c12a188d95ec2a6c99dfe5dc488e/docs/tutorials/embedding/CovariateAssistedEmbedding.ipynb

Closes #134

[netlify]

Deploy Preview for graspologic failed.

Built with commit ab13228

https://app.netlify.com/sites/graspologic/deploys/60a6ae9221c0bc0007ac05c2

[loftusa]

@bdpedigo Ok, I think code is working well enough to look at it? still a bit early and there's a lot of debugging code floating around, but feel free to take a first pass if you'd like.

a few things:

• I implemented the tuning algorithm over the course of the last few days
• Code for grabbing the tuning parameter is still pretty messy, I'm planning to clean it up
• I currently save the laplacian, the covariates, and a few functions of them into the class instantiation object in fit. That's somewhat temporary, planning on changing that as well
• the tuning parameter algorithm needs to fit within a range a bunch of times and then run k-means repeatedly (then return whichever one has the smallest k-means inertia). To do that, I need to re-embed repeatedly, but I don't want to call self.reduce_dim_ a bunch. Just wrote my own code for it, but feels like it's repeating code from the superclass a bit
• haven't made the updated figure yet (it's gonna take a long time to run)
• on that note, current iteration of the tuning algorithm takes about 10 seconds to fit on a 300-node graph, since it has to run k-means in a loop over a tuning space. Can't really think of a way to speed it up.
• tests broke since I changed some stuff after implementing the tuning algorithm, shouldn't take too long to get them up and running again though

[bdpedigo]

@loftusa mind resolving any comments you've addressed and need no further discussion?

[loftusa]

@loftusa mind resolving any comments you've addressed and need no further discussion?

done

[bdpedigo]

just a heads up that tests aren't passing anymore

[loftusa]

just a heads up that tests aren't passing anymore

fixd

[bdpedigo]

what did you end up doing regarding our conversation about checking for directed graphs?

[loftusa]

what did you end up doing regarding our conversation about checking for directed graphs?

#599 (comment)

[bdpedigo]

@loftusa no rush on my end but can you just re-request review when tests are passing and everything looks good to go?

[loftusa]

@loftusa no rush on my end but can you just re-request review when tests are passing and everything looks good to go?

ok, tests are passing, except for the deploy preview, but it looks like that's some issue with the new graspologic-native version that got released two days ago rather than anything in this PR

(maybe I'm wrong?)

[bdpedigo]

thanks @loftusa 🎉

[loftusa]

thanks @loftusa 🎉

Woo!