fit_transform hangs for low number of samples

[fredmell]

Hello,

Thank you for this great package! It seems like I have run into a bug with the most recent version.

fit_transform hangs indefinitely for n=30 and less. It runs fine for values larger than this.

import numpy as np
import pacmap 

n = 20
X = np.random.normal(size=(n, 512))

embedding = pacmap.PaCMAP(n_components=2) 

X_transformed = embedding.fit_transform(X)

print(X_transformed.shape)

It seems to have been introduced in 0.7.4, it runs to completion on 0.7.3.

Let me know if there is anything more I can provide.

[hyhuang00]

Thanks for reporting. I've replicated this error. I suspect this is a similar issue like #72 and will work on this as soon as possible.

[hadronzoo]

This might be caused by the sample_FP function entering an infinite loop if n_samples > maximum - len(reject_ind) - !is_in(self_ind, reject_ind).

[hyhuang00]

Resolved in release 0.7.6. Please reopen this issue if you encountered this issue again.