• thevoidzero@lemmy.world
    link
    fedilink
    English
    arrow-up
    2
    ·
    1 month ago

    It’d be easier to do 9e5, 8e5 and so on though. Linear decrease in time with each optimization. 1e5.9 seems risky.

    • Iron Lynx@lemmy.world
      link
      fedilink
      English
      arrow-up
      1
      ·
      1 month ago

      Depends on how and/or if you want to curve future changes. Going 1e5.9, 1e5.8, 1e5.7, … will curve logarithmically, while 9e5, 8e5, 7e5, … will curve linearly within each power of ten, then get a discontinuity at 1e5 and go 9e4 and scale linearly at a different rate.

      Of course, you’ll have to be an absolute nerd to find that a problem and there’s a nonzero chance that I’m such a nerd and I just admitted my guilt 😅