AlphaGo's memory can be messed up when against a random number generator

送交者: conner 于 2016-03-15, 09:21:14:

AlphaGo memorizes the route analysis and use that as the basis of the next round analysis. When playing against a random number generator, the basis will be rebuilt for each move.

所有跟贴:

蠢货，虽然你不能用alphago 但是可以去那其他围棋软件试试看。 (无内容) - bluesea (0 bytes) 2016-03-15, 21:26:31 (803996)
it summarize patterns to evaluate the cost. not based on sample route. (无内容) - 短江学者 (0 bytes) 2016-03-15, 10:00:16 (803957)
- The searching tree continues to grow and go deep. if trimmed, will lose power (无内容) - conner (0 bytes) 2016-03-15, 10:40:25 (803962)
  - by deep I mean deep with a high belief bayesian probability (无内容) - conner (0 bytes) 2016-03-15, 10:58:59 (803967)
  - yes but after a while it shrinks -an extreme is just like 中盘投子。 (无内容) - 短江学者 (0 bytes) 2016-03-15, 10:42:36 (803964)
    - yes, when the tree is complete in the depth, the result is decided - conner (50 bytes) 2016-03-15, 10:45:52 (803966)
- Q-learning is approximate dynamic programming with expectation - 短江学者 (169 bytes) 2016-03-15, 10:05:01 (803958)
u think it erased the cache? what kind of bug will lead to that (无内容) - pgss (0 bytes) 2016-03-15, 09:57:19 (803955)
- not a bug. when the move is not on the overlapped routes of the search tree - conner (124 bytes) 2016-03-15, 10:37:53 (803961)
how does a computer memorize stuff? - pgss (80 bytes) 2016-03-15, 09:37:27 (803952)
- memorization is widely used programming technique - conner (113 bytes) 2016-03-15, 09:40:04 (803954)

加跟贴