Sean (Nov 24 2020 at 18:30): Sean (Nov 24 2020 at 18:30):@starseeker I was working on spherical sampling and noticed you added a sobol sequence generator and related code in libanalyze. That's awesome! Is that exposed/presented anywhere user-visible?
Sean (Nov 24 2020 at 18:31):I want to compare it with another generator I have for incremental spherical sets, use it as a base for convergence testing on closures.
starseeker (Nov 24 2020 at 18:55):It's exposed in the pnts command, IIRC... one sec
starseeker (Nov 24 2020 at 18:57):pnts gen --sobol
starseeker (Nov 24 2020 at 18:57):If you generate points from a sphere using --rand and --sobol, you'll see the difference
starseeker (Nov 24 2020 at 18:58):sobol ended up being of dubious utility for that - it clusters near the poles of the sphere, and I never got deep enough in to see if there was a correction that might be applied.
starseeker (Nov 24 2020 at 19:00):Really should have ripped it out, to be honest - it took a while to find a sobol generator that was appropriately licensed and able to be integrated cleanly, so I hated to get rid of it, but honesty compels me to admit it's mostly code bloat right now...
Sean (Nov 24 2020 at 19:02):clustering near the poles is just a classic mistake in how the random numbers are being used
Sean (Nov 24 2020 at 19:02):the code itself is simple enough and sobol is a useful sequence for quasi-random montecarlo
Sean (Nov 24 2020 at 19:05):I did see a few issues in the code that need clean up, few mistakes to correct
Sean (Nov 24 2020 at 19:06):few things I just don't understand yet (like the next ranges and skip function, the example snippet has typos and comments aren't clear to me yet)
Sean (Nov 24 2020 at 19:08):thanks for the lead on pnts. I'll check it out.
Last updated: Oct 09 2024 at 00:44 UTC