search

pierre-guillou / pdiags_bench

3 stars 1 forks source link

Updates #2

Open mglisse opened 1 year ago

mglisse commented 1 year ago

Hello,

I don't know if you intend to maintain the benchmark now that your paper is out, but just in case:

  1. Could you try the version of CubicalRipser at https://github.com/shizuo-kaji/CubicalRipser_3dim? It may require minor tweaks to run on recent python (see the github issue there), but nothing complicated.
  2. For Gudhi, I would recommend CubicalPersistence rather than CubicalComplex. Currently, this only matters in 1D, but it may matter more in the future.
  3. The paper puts Gudhi in the parallel implementations, but Gudhi doesn't really use parallelism for persistence (just a parallel_sort where a plain sort would do). For comparisons, it would fit better with the other single-threaded implementations, possibly configuring it with -DWITH_GUDHI_USE_TBB=OFF to make sure even sorting does not use threads.
pierre-guillou commented 1 year ago

Thanks for your interest in this project! Sadly, I changed jobs a few months ago and I have less time to work on it. I'll try to find some time to look at this issue next month during my summer break.