dvonnteam team mailing list archive

[noreply@xxxxxxxxxxxxx]

------------------------------------------------------------
revno: 24
author: Louis Reymondin
committer: louis-reymondin
branch nick: trunk
timestamp: Wed 2011-09-14 16:05:22 -0500
message:
  I cleaned a bit Board.cpp and implemented a function which approximates the likelihood for a cell to have a cross or a circle (really basic and certainly inaccurate but let s start with something).
  
  fCellPlayerPresence will return 1 if the cross/circle is in a collapsed state, otherwise it will return 
  
  (Let s say that the color is circle) 
  Sum for each circle presents in the cell (1-1/(2*Nc))/N
  
  Nc: number of elements (circle + cross) in the conjugate cell
  N: number of elements (circle + cross) in the evaluated cell
  
  I then changed the SaveBoard and SaveOutput function to keep the stat for the whole board. I ll try to use this to train basic neural networks which take as inputs the values of fCellPlayerPresence for circle and cross (32 inputs) and which try to predict the score 
  
  This is still a basic approach but I m keen to see the results
modified:
  Board.cpp
  Board.h


--
lp:qttt
https://code.launchpad.net/~dvonnteam/qttt/trunk

Your team DvonnTeam is subscribed to branch lp:qttt.
To unsubscribe from this branch go to https://code.launchpad.net/~dvonnteam/qttt/trunk/+edit-subscription