You do not give details on the indexes chosen in the comparison methods. From the runtime numbers, it appears that you did not add enable an index in ELKI? Do the runtimes include JVM startup cost, R startup cost, or did you use internal runtime measurements (e.g. -time).
https://rdrr.io/cran/dbscan/f/inst/doc/dbscan.pdf

I suggest to make separate benchmarks with and without indexing, and

Note that this is different from #60, as that one concerns only scaling.

Here are some results from TIOGA (with CudaUVM). Three variants:

• Upstream master
• Using unsigned int for size_type template parameter in Kokkos' BinSort [here]
• Using Thrust [[here](http

Provide some detailed information for real-time algorithm processing and about how it works. Maybe open a wiki for topics.

This issue created from issue #8