Knuth Releases Part Of Volume 4 31
Grendel Drago writes: "Donald E. Knuth has released "Pre-Fascicle 2b: Generating all permutations" from TAOCP Volume 4. It will be section 7.2.1.2 of the final work. Oh, and Volume 4 may now fill *four* subvolumes. Send in bugs, get checks for $2.56, tell the grandkids."
from the inside flap... (Score:2, Interesting)
From amazon [amazon.com].
TAOCP has got to be one of the greatest examples of a genius mind at work. Knuth's work should be placed in the same echelon as Shakespear, Chaucer, and Hemingway.
-Vic
More interesting... (Score:3, Interesting)
Evidently the random number generator ran_array / RNARRAY from Volume 2 had some problems. As I read it, if one seed is used many times, it would produce numbers that passed randomness tests; but one user tried many different seeds for only a few generations - which began to fail randomness tests.
The remarkable thing Knuth noted was that two different methods of fixing it were found by Richard Brent. The first was to discard the first 2000 numbers; the other was just basic improvement of the initialization of the algorithm.
I'm very curious as to why this is; my understanding of seminumerical theory is limited to what I've read in Knuth, but I'm still very interested in the causes of this problem.