Identify the empty droplets from the per-barcode library sizes

Kept on the Rust side so the knee detector exists once: it has to match scipy.ndimage.gaussian_filter1d and numpy.gradient closely enough to land on the same rank as the CellSweep reference, and a second copy in R would drift.

Usage

rs_sc_infer_empty_droplets(lib_size, empty_params)

Arguments

lib_size

Integer vector. Library size per barcode, in store order.

empty_params

List. Parameter list, see params_sc_empty_droplets(). "supplied" is rejected, since there is nothing to infer.

Value

A logical vector that is TRUE where the barcode is an empty droplet.