From: kingpin@freemail.it   
      
   Hi,   
   i want to write a program that solve the peg solitaire game.   
   The goal is to remove pegs from the board by jumping over each peg   
   with another peg; this removes the "jumped" peg (similar to Checkers   
   jumps). Only horizontal and vertical jumps are allowed. The game is   
   over when no more jumps are possible. You win the game by removing all   
   the pegs except one from the board. For example, if the board is 4x4,   
   a possible configuration is:   
      
   . o . .   
   . o . .   
   . . . .   
   . . . .   
      
   and the solution is   
      
   . . . .   
   . . . .   
   . o . .   
   . . . .   
      
      
      
   For large boards (7x7...), a simple brute-force search algorithm take   
   too much to solve the game, given an arbitrary configuration.   
      
   So i was thinking about neural networks. I would like to know if a   
   neural   
   network can be used to say if a given configuration can be solved (or   
   to give the probability that the configuration can be solved).   
      
   In this way, i could order the sequence of possible moves using the   
   net.   
   This should improve the speed of the search.   
      
   I don't know much of neural networks. I've tried to write such a   
   program, but   
   it doesn't function. I don't really know if this is a problem of my   
   code or is the function itself that cannot be learned. The problem is   
   that   
   the training phase does not converge to a low error rate.   
   Maybe this function is too chaotic to be learned, or maybe i must wait   
   2 day   
   to get a low error rate.   
   I *really* don't know!   
      
   Thanks,   
   Lucas   
      
   [ comp.ai is moderated.  To submit, just post and be patient, or if ]   
   [ that fails mail your article to , and ]   
   [ ask your news administrator to fix the problems with your system. ]   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)