BGonline.org Forums

XG training and NN question

Posted By: Klaus Evers
Date: Sunday, 1 August 2010, at 4:30 a.m.

In Response To: XG training and NN question (Timothy Chow)

However, perhaps there's some way of training two separate neural nets, the "normal" NN and the "deep backgame" NN, and then have the bot choose which NN to use (perhaps based on a third NN?).

From what I know, this approach is used by SNOWIE, with seven (not two) NN and another one that decides which to choose for a given position.

And this (two-step process) is one of the reasons that SNOWIE is so slow (compared to today's standards)

Messages In This Thread

• XG training and NN question
  Leonardo Jerkovic -- Sunday, 1 August 2010, at 2:31 a.m.
  • XG training and NN question
    eXtreme Gammon -- Sunday, 1 August 2010, at 3:06 a.m.
    • XG training and NN question
      Timothy Chow -- Sunday, 1 August 2010, at 3:31 a.m.
      • XG training and NN question
        Klaus Evers -- Sunday, 1 August 2010, at 4:30 a.m.
      • XG training and NN question
        Tom Keith -- Sunday, 1 August 2010, at 1:48 p.m.
        • XG training and NN question
          Timothy Chow -- Sunday, 1 August 2010, at 5:33 p.m.
          • XG training and NN question
            Leonardo Jerkovic -- Sunday, 1 August 2010, at 6:27 p.m.
            • XG training and NN question
              Timothy Chow -- Sunday, 1 August 2010, at 10:43 p.m.
          • XG training and NN question
            Tom Keith -- Sunday, 1 August 2010, at 6:53 p.m.
            • XG training and NN question
              Timothy Chow -- Sunday, 1 August 2010, at 10:46 p.m.
              • XG training and NN question
                mtuhtan -- Sunday, 1 August 2010, at 11:58 p.m.
                • XG training and NN question
                  Leonardo Jerkovic -- Monday, 2 August 2010, at 12:33 a.m.
    • XG training and NN question
      neilkaz -- Sunday, 1 August 2010, at 3:36 a.m.
    • V2 feature request
      KenB -- Sunday, 1 August 2010, at 6:09 p.m.