[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

RE: VMs: superblock update

Bruce  [re: genetic algorithms]

Let me recap a little to explain what I'm doing.

Having found many families of frequent repeated sequences which 
vary only slightly from each other, I tried to put the 'pieces' 
together into a block.  This was done by placing the pieces onto
an empty board, trying them in order of decreasing frequency but
each time placing the 'piece' which fitted best with what was
already on the board.  I also used a table of the valid 5-char
sequences to check that each new placing wasn't introducing 
5-char sequences to the table which do not occur in the VMs 
itself.  This is not guaranteed to be minimal but works pretty 

The rules for using a block is simply to move a pointer through 
the manuscript and at each position finding the longest match 
contained in the block (with a minimum of 6 chars).  If no 
acceptable match is found you move the pointer on one place and
try again.  There is no upper limit on the match length.

I have a block of 1500 chars which can produce 58% of the VMs,
and a bigger one of 7000 chars which can do 72%.

I do have a fascination with genetic algorithms; I once wrote a
program to compress a text file genetically, and another one I 
did tried to solve the 'eternity' puzzle.   The concept and 
method is certainly appropriate to finding a minimal block, or 
evaluating a given block, but both are problems in many thousands
of variables and as such you'd need access to a LOT of computing 
resources and time to use effectively. :-(


To unsubscribe, send mail to majordomo@xxxxxxxxxxx with a body saying:
unsubscribe vms-list