HYPEX(1)

📦 version 1.14, in nixos 26.05, nixpkgs unstable

💅 HYPEX(1) — the hypex girlie's field guide ✨

okay so bestie, welcome to the hypex page — manual section 1, the big leagues. stay with me, this one's niche but it slaps 🎀

✨ NAME ✨

hypex — computes the Chernoff exponent between two simple categories. that's it, that's the whole personality. 👑

🌸 SYNOPSIS 🌸

hypex [-hH size] [-s stepsize] CATDUMP1 CATDUMP2

hmine -V

(yes, hmine -V is literally what's printed there. no, we don't ask questions. 🫶)

🦋 DESCRIPTION 🦋

so hypex reads two category dumps that dbacl(1) produced after learning. a category dump comes from using the -d and -l switches, and it's basically a textual representation of the feature weights that live inside the binary category files. like, the binary files but in a form you can actually read. iconic of them honestly 💖

give it two of those dumps for simple unigram categories and hypex will:

  • calculate the Kullback Leibler divergence between the probability models, and
  • print out exponential error exponents for Neyman-Pearson hypothesis tests, across a whole range of threshold values.

for the lore, see Cover and Thomas (1991) Elements of Information Theory, Chap. 12. 📚

heads up bestie: only simple categories are supported, and the output only makes sense under appropriate theoretical conditions. this is a very specialized calculator. like, do not bring this to a general-purpose party. 🙈

⭐ EXIT STATUS ⭐

  • 0 — success, we ate 💅
  • 1 — an error occurred, crying 😭

💜 OPTIONS 💜

  • -h — it's listed in the manual with no description attached. mysterious. main character energy, zero elaboration. 🕵️‍♀️
  • -H — same as dbacl(1). selects the hash sizes in powers of two. ✨
  • -s — stepsize for the threshold. hypex outputs exponents for different values of the threshold, within an interval bounded by the Kullback Leibler divergences between the categories. yes, that's a whole sentence, and yes, it's exactly what it means. 🎀
  • -V — print the program version number and exit. bye bestie 👋

🌈 SOURCE 🌈

the source code for the latest version of this program is available at the following locations, go get her:

http://www.lbreyer.com/gpl.html
http://dbacl.sourceforge.net

💀 BUGS 💀

  • doesn't work with complex categories. nope. not happening. 🚫
  • and the theoretical assumptions are unrealistic in practice. so like... the math is gorgeous, the real world is not. 🤡

👑 AUTHOR 👑

Laird A. Breyer <laird@lbreyer.com> — the icon, the legend, the person who wrote this and dipped 🫶

🍓 SEE ALSO 🍓

dbacl(1), mailcross(1), mailfoot(1), mailinspect(1), mailtoe(1), regex(7) — the whole friend group, go say hi 🌸

and that's the page bb!! version 1.14, from the Bayesian Classification Tools era, HYPEX(1) forever — mwah, love u, go compute some exponents ✨💖🎀

📦 where this page comes from