On 5/12/07, Chris Fant <[EMAIL PROTECTED]> wrote:
> John corrected me. It turns out we do add the playouts from the
> possible moves (we didn't used to in my original implementation, but
> he changed that). The difference with what Jason described is that we
> do not use the playout count from the destination node. Instead, we
> keep counters at the moves, which means that we don't use the "1000"
> at all. Anyway, that doesn't change the essence of what I said much.
I did the same thing you did as far as the transposition table
implementation (UCT tree with pointers into the transposition table).
But I don't understand why you are using a counter in the tree as
opposed to a counter in the transposition table. Yes, the fact that a
node has been found in the transposition table and has many
simulations already tried from it will mean that it will not get
explored as much as an unseen node will (it will still get exploited
if it has performed well). But that is correct behavior. It does not
need to be explored as much as a node that was not found in the table.
You might be right, so I'll give this some more thought. I'll put this
in my to-do list of experiments.
Álvaro.
___
computer-go mailing list
computer-go@computer-go.org
http://www.computer-go.org/mailman/listinfo/computer-go/
