On May 11, 2010, at 6:35 AM, David Fotland wrote:

> It is quite memory hungry since near the leaves there are many positions
> with very few active children.  On 9x9 I let a few playouts go through a
> move before expanding it, and that gives me plenty of memory.  I ship with a
> default of 300 MB reserved for the hash table, and that's enough.

So that confirms more or less how I thought hash-tables were implemented 
(although others may still use different methods of course). Using rather more 
memory than a tree implementation. But I see the advantages in terms of 
efficiency.

What I do in the tree is when I first expand a node I stick in a small 
place-holder that just has the move and the RAVE values (and a bit more) but no 
child information. Only when a node gets selected because it has the best RAVE 
score does it get replaced by a full node. On average less than 10% of the 
nodes get replaced by a full node.

You could probably do something similar with a hash-table. But the 
implementation might get a little more complex.

Mark

_______________________________________________
Computer-go mailing list
[email protected]
http://dvandva.org/cgi-bin/mailman/listinfo/computer-go

Reply via email to