BGonline.org Forums

Horizon effect?

Posted By: Maik Stiebler
Date: Tuesday, 15 March 2016, at 6:39 p.m.

In Response To: Horizon effect? (Sam Pottle)

That makes a lot of sense. In today's stream Michael Redmond discussed that issue (at least I think so, as he did not use the actual term 'horizon effect'), talking about how it is a notorious problem in previous Monte Carlo tree search based Go programs, how some programmers reading DeepMind's Nature paper thought that there was nothing in there that indicated that the problem had been solved and how the DeepMind programmers seem to be uncertain of it being the explanation for the things we saw on Sunday, one of them telling Redmond that "it had more to do with the neural networks".

Messages In This Thread

• AlphaGo - Lee Sedol Spoiler (OT)
  Casper Van der Tak -- Sunday, 13 March 2016, at 8:48 a.m.
  • AlphaGo - Lee Sedol Spoiler (OT)
    Taper_Mike -- Sunday, 13 March 2016, at 10:22 a.m.
    • AlphaGo - Lee Sedol Spoiler (OT)
      Maik Stiebler -- Monday, 14 March 2016, at 7:54 a.m.
      • Horizon effect?
        Sam Pottle -- Monday, 14 March 2016, at 5:18 p.m.
        • Horizon effect?
          Maik Stiebler -- Tuesday, 15 March 2016, at 6:39 p.m.
          • Horizon effect?
            Taper_Mike -- Wednesday, 16 March 2016, at 9:13 p.m.