return model with precalculated nearest neighbor

[yuhanH]

Hi,
I noticed that umap cannot return a model with precalculated nearest neighbor input in nn_method.

I would propose that modifications of two parts will enable umap to return the model with precalculated nearest neighbor input.

part1:

uwot/R/uwot.R

Line 1656 in cbb78b5

scale_info = attr_to_scale_info(X),

scale_info = if (!is.null(X)) { attr_to_scale_info(X) } else { NULL },

part2:

uwot/R/uwot.R

Line 1687 in cbb78b5

res$metric[[1]] <- list(ndim = length(res$nn_index$getItemsVector(0)))

    if(!is.null(X)){
            res$metric[[1]] <- list(ndim = length(res$nn_index$getItemsVector(0)))
          } else{
            res$metric[[1]] <- list()
          }

Happy to put a PR for this if people agree.

[jlmelville]

Hi @yuhanH, uwot doesn't return a model when using precalculated nearest neighbors, because you can't transform new points with just that information. If you think there is a use for the model without that I would be happy to look at a PR, but can you explain a use case?

[jlmelville]

Closed by #64