OpenEuphoria

Re: $100 Contest Question

• Posted by David Cuny <dcuny at LANSET.COM> Mar 05, 2002
• 461 views

Martin wrote:

> Maybe I should move to #3 but I do not dare to 
> challenge David Cuny.

I'm not the one to worry about. Doesn't C. K. Lester have his "White Rabbit" 
program solving crytograms in about 5 seconds? I'm using a fairly naive 
genetic algorithm (GA) as the core of my solver. Even with a number of 
optimizations, it took about an hour to crunch through:

  king nero prvby khnc avcl fhyyvy cavrdlk chfn
  deaf tfengy ic dhuly erbv ychtl
  riby xhff ertfikerd fhyyel

before finally settling on:

  dumb minx knots damp rope lassos pronged palm
  girl climbs up gazes into space
  nuts fall including lassie

There may be room in the code for improvement - profiling will tell - but a 
10:1 improvement? I suspect not, especially when "White Rabbit" is apparently 
churning out the solution in seconds.

The real problem for me seems to be the algorithm. GA is dead simple to 
implement, and will eventually hit on the solution. But it's not especially 
efficient. I'm considering using trigraphs instead; they look *very* 
successful. Take a look at:

   http://www.muth.org/Robert/Cipher/substitution

Because the topic of solving crytograms is so well researched, I suspect the 
winners will all select the same general algorithm. So the prize will go to 
the person who bothers to do a little research, and optimizes the heck out of 
their code. That sounds like jiri, C.K. Lester or  Derek... but certainly not 
me. I'll don't take my code down to that sort of optimization.

Thanks anyway for the vote!

-- David Cuny