Heavy playouts (was Re: [computer-go] Efficiently selecting a point to play in a random playout)

[Peter Drake]

Let me mildly retract this and say that method 2 appears to be better  
than method 1.

Peter Drake
http://www.lclark.edu/~drake/



On Jun 6, 2007, at 2:51 PM, Peter Drake wrote:

> I assume you mean "...than finding tricks to accelerate RANDOM  
> playouts".
>
> I've been messing around with where to apply heuristics. Candidates  
> include:
>
> 1) In the UCT tree, since this is earliest in each playout.
> 2) In the moves beyond the tree ("heavy playouts"), since this is  
> where most of the moves happen. Because this part is so speed- 
> critical, some trick often has to be used so that you don't  
> evaluate every possible move on every simulated turn. Options include:
> 	2a) Pick N possible moves and choose the one the heuristic likes  
> best. (N might be 3 or so.)
> 	2b) Only apply the heuristic on the first N moves beyond the  
> fringe of the tree.
>
> It currently looks like 2b is by far the best place to apply  
> heuristics. Do others have conflicting results?
>
> Peter Drake
> http://www.lclark.edu/~drake/
>
>
>
> On Jun 6, 2007, at 2:41 PM, Rémi Coulom wrote:
>
>> I am sure it could be made significantly faster, but adding  
>> knowledge and high-level algorithmic improvements is tremendously  
>> more profitable in terms of playing strength than finding tricks  
>> to accelerate playouts.
>>
>> Rémi
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://computer-go.org/pipermail/computer-go/attachments/20070606/b8fcfe2a/attachment.htm