Wiener-san no chobits

I finished watching "Chobits" disc six yesterday, and realized that distributional ranking is not only a mash-up between the gist of Knuth volume two and volume three, but a simple combination of (ideas from, to be specific) two of last century's "greatest" algorithms.

I wasn't sure what to think of "Chobits" at first -- it comes from CLAMP after all -- but it got better by disc five or so (go Sumono!), and watching these episodes on disc six, I thought about how how Hideki : Chi :: Knuth : 6502 650.

I'm guessing Knuth doesn't watch anime, but perhaps if he did, he'd like these few episodes of "Chobits". Or, maybe not.

In Knuth vol. 2, Knuth writes extensively on "semi-numerical algorithms", like but not such as the Mersenne Twister. In vol. 3, he surveys "sorting and searching" -- two themes that I used to think were the heart of 20th century computer science.

I think I'm starting to realize now the importance of semi-numerics. Not only in pseudo-random number generation, but in quasi-randomness and low-discrepancy sampling theory.

Given that, it seems ambitious and yet plausible to claim that distributional ranking combines semi-numerics with the combinatorial algorithmics of classic "sort and search".

How? Well, of the ten "greatest" algorithms of the 20th century, I see the Monte Carlo method as the brainy essence of semi-numerical analysis (scientific computing) and "qsort" (randomized sorting) as the heart-y core of combinatorial algorithmics.

What I realized subconsciously while watching "Chobits" was that one could see classic Monte Carlo and qsort as special cases of donkey sort and (more generally) triple-you ranking.

Monte Carlo integration, after all, could be considered a distributional algorithm that throws away the combinatorial ordering that donkey sort provides, and simply returns an arithmetic sum.

Quick sort, then, could also be considered a distributional algorithm run with a trivial, not-so-pseudo random number sequence sampler.

This is all and well-seemingly twisty good, but what can we do with this possible knowledge? For one, let's try to improve how we humans (and computers?) self-schedule, in our next entry, to be titled, "I, Banking"...

---

After I wrote this yesterday, I realized that I know a little bit about Monte Carlo (by way of one of Gene's students), even if I never took that class on randomized algorithms from Dr. Blabla, but I don't know that much about "Metropolis" itself.

I also skimmed through biographies of Wiener, Turing, and Nash yesterday, which gave me the impression that (one) this von Neumann fellow is a lil sketchy and (two) I seem to identify more with Wiener-Einstein than with Turing-Nash, for some reason or another.

I'm going to have to read Wiener's work more carefully now, dense as it may be.

If triple-you combines basic ideas from algorithms (1) and (7) on the top ten list, then I'm going to conjecture that, once I have a better grasp on how algorithms (10) and (3) work, I might be able to code something really interesting :)