1. 73

  2. 12

    This “try multiple algorithms until one of them happens to work” approach is profoundly unethical — especially since we don’t have separate training, test, and validation sets — but at least we’re being honest about what we’re doing, instead of inventing a fancy-but-obfuscatory technical term like “ensemble methods” or “hyperparameter tuning”.

    I really enjoy the spirit of SIGBOVIK :)

    1. 1

      God every time I read ensemble methods in a research paper grr

    2. 6
      1. 4

        And people still say perl is ugly.

        1. 2

          A follow-up study should do the same with Perl6 specifically and compare the results.

          1. 2

            None of those strings are valid Perl6 programs.