-
Notifications
You must be signed in to change notification settings - Fork 1
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
Replace existing port tests with automated regression tests #42
Comments
I started experimenting a bit with
results in this serialization:
whereas this test:
results in this snipped serialization:
|
Wow, that default serialization for arrays looks horribly inefficient! I like your solution, although if we can avoid having to manually serialize altogether by going with a different tool, I'm definitely open to that. I can't remember why now, but I also got the impression that |
@aazuspan, I've been playing around with ps. At the risk of overburdening you, please feel free to weigh in on any issues or PRs that I'm putting up over in synthetic-knn. I'll try to be mostly independent on this one, unless it piques your interest and then I'd be overjoyed if you want to jump in 😄. |
That's awesome @grovduck! I took a quick look and the regression test looks remarkably simple to implement!
Interesting, so we would still be using the
I'm very curious, so I'll keep an eye on development over there! |
Sorry, that wasn't very clear. I think we'd want to do the following:
There's a chance that we'll need to go back to creating other files through |
Ah, that makes perfect sense! Thanks for laying out the game plan. |
Currently, we test our estimator accuracy against manually generated results from
yaImpute
andpynnmap
. Once we have all major functionality implemented and are confident in our results, we can switch to regression testing to ensure no errors are introduced. This was briefly discussed in #40:We'll need to do some experimenting to find the right tool for our use case, but the advantage should be a massive simplification of our testing system (e.g. the removal of
KNNTestDataset
) and the ability to easily test against a wider range of parameters (e.g. values ofn_neighbors
andn_components
).The text was updated successfully, but these errors were encountered: