Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: algorithm question

Author: Dann Corbit

Date: 13:05:04 03/01/06

Go up one level in this thread


On February 28, 2006 at 20:38:08, Ratko V Tomic wrote:

>> Could you please e-mail me, as I think this has
>> wandered off topic for CCC.
>
>I just updated my CCC email in the profile (the old one
>was receiving too much spam).
>
>> I'm intrigued as to how you think this could be utilised
>> for compression algorithms.
>
>This algorithm would fall under combinatorial generation
>(all subsets of a subset). There is neat online book by
>Ruskey with lots of related, practically oriented material:
>
>    F. Ruskey "Combinatorial Generation"
>    CSC-425/520 Univ. of Victoria CA,2003
>    http://www.1stworks.com/ref/RuskeyCombGen.pdf
>    http://www.cs.uvic.ca/~ruskey/Publications/
>
>See also [8a] (Brak's AMSI lectures), [21] (Knuth4 on bits &
>boolean manipulations), and [25] (Ruskey on permutations) on the
>QI Ref. Lib. page: http://www.1stworks.com/ref/RefLib.htm .
>There is also interesting series of papers on arXiv which uses bit
>pattern manipulations to compute roots of integers & polynomials:
>http://arxiv.org/find/math/1/au:+Mittal_A/0/1/0/all/0/1 .
>
>I have used similar 'carry ripple across' techniques many times,
>but your variant is a particularly elegant one which I
>didn't see before (and it didn't occur to me). It is clear
>that one can extend it to bit arrays/sets of any size. One
>possibility for using it in compression might be as an
>interpolation for the ranking. In combinatorial compression
>you compute rank (eg. lexicographic) of a given specific bit
>pattern within the larger set of patterns satisfying some
>constraints (e.g. specified bit counts or number of runs etc).
>But you can also compute rank for a set of bit patterns (which
>may be easier or use smaller tables). The subset generation
>would then be used to interpolate witin the decoded smaller
>set of patterns.

I agree that this simple, elegant algorithm is a significant, mathematical
discovery.

I think it would be good to let people know about it.  It might be worthwhile to
write a paper on the algorithm.  Even a one-pager, just so people can find out
about it when it gets into the Citeseer database.



This page took 0 seconds to execute

Last modified: Thu, 15 Apr 21 08:11:13 -0700

Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.