James Melville

>In the case you mentioned, where nearest neighbor descent tends to repeat distance calculations a lot, is there a value, e.g. what faction of the dataset distance, was used to...

This feature is available, but only if installing directly from this repo. Otherwise, it will become available when there is a new release of this package.

The PCA initialization option isn't in the 0.5.3 release (which came out last year).

The most recent release is 0.5.3. If you are looking at that seems to refer to the current state of the `master` branch of this repo. If you want to...

More likely that core dependencies like numba didn't support 3.11 at the time of 3.11's release.

Using precomputed nearest neighbors is covered at and . Pearson correlation is now supported with `metric = "correlation"`.

Other things that could change: * `init_sdev = "range"` by default. * if `batch = TRUE`, we can set `n_sgd_threads` to the same value as `n_threads` * I almost always...

#123 adds these changes, including `umap2`. Also `batch = TRUE` is a default. I am likely to set this is version 0.2 when submitted to CRAN.

I agree some kind of diagnostic plotting is necessary for any dimensionality method which embeds a neighbor graph. I have written substantial amounts of R (and Python) plotting code for...

> E.g. 2 clusters may sit equally close to a third cluster but only one of them is close due to contentedness, thus meaningful, the other may only end up...