[computer-go] UCT

[Peter Drake]

On Nov 3, 2006, at 8:33 AM, Rémi Coulom wrote:

> I have never had memory problems with Crazy Stone. Crazy Stone uses  
> 32 bytes of RAM per node of the tree. It is not really a tree, but  
> a hash table, to handle transpositions. For nodes closer to the  
> root, I "extend" each node with more data (pointers to children).  
> This is only

Since Orego stores the entire tree (with some optimizations for non- 
branching paths), it is ... somewhat less memory-efficient than Crazy  
Stone.  :-)

> needed for a small proportion of the total number of nodes, so it  
> does not eat up a huge amount of RAM. That is because I don't do  
> UCT in a node before 81 random simulations have been run there.

A thought: if we define the uncertainty of an untried move to be 2,  
the UCT algorithm will automatically try all of those before trying  
anything else twice.

Maybe you're already doing that.

> Since I use transpositions, it is not really possible to remove a  
> subtree, anyways.

Yes, another thing I have to (re-)implement...

Peter Drake
Assistant Professor of Computer Science
Lewis & Clark College
http://www.lclark.edu/~drake/