Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Hash codes

Author: Ralf Elvsén

Date: 01:48:05 11/14/01

Go up one level in this thread


On November 13, 2001 at 19:19:46, David Hanley wrote:

>It's my understanding that the hash codes used in a chess program are hamming
>codes, and they are generated such that
>
>bit_count( code[x]^ code[y] ) > threshold
>
>for all combinations of codes, obviously self excluded.
>
>Is this correct?
>
>I plan on using 48 bit hash codes, and my threshold for bit mismatches is
>hash_length/6, or in the case of 48 bit keys, 8 bit differences.  Is this
>sufficient?
>

It has been argued here (and I think it is correct) that the Hamming
distance shouldn't be too big either, i.e. 47 is as bad as 1.

I did this for 64 bits, and I tried to push all pairwise Hamming
distances as close to 32 as possible, e.g.

minimize( maximum(abs(32 - Hammingdistance(r_i, r_j)) ) )

where r_i are the random numbers and i != j . I got a value of 24 or
so (don't remember exactly). It seems to work fairly well.

Ralf



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.