James Melville

I failed to make any progress this weekend, but it is closer to the top of my to-do pile.

Sorry I have made zero contributions to this so far. I was traveling for the last two weeks and had little to no internet access.

> Close enough, I'd say. I know we're identical up to the optimization, so I'm guessing that the differences are due to our different PRNGs - I'm using `std::mt19937_64` to...

If you have a model generated by e.g.: ```R iris_umap

On the `master` branch, `min_dist` and `spread` are now exported as part of the model when `ret_model = TRUE`.

The description in the doc for `?dist` seems to be indicate that `method = "binary"` is the Jaccard/Tanimoto distance. Supporting this will require an alternative to Annoy, so it won't...

Currently I have no plans to add densMAP. The implementation would require adding extra routines to the generic optimization code which would in principle affect all other methods, so the...

Update: I still have no plans to add densMAP to uwot. *But* I have been experimenting with an approximation to it which seems... fine. Instead of adding an extra objective...

On the master branch the promised approximation to densMAP is now available. Set `dens_scale= 1` when you call the `umap` function. The closer that number goes to `0` the closer...

> I also see that leopold does a better job where densMap does not. I wouldn't claim that leopold does better than densMAP. I expect densMAP will always do better...