Re: [Bug-gnubg] Question about higher plies

[Joern Thyssen]

On Mon, Jul 26, 2004 at 03:03:00PM +0200, Joachim Matussek wrote
> Hello,
> 
> I was musing a little bit and following question arose.
> 
> E.g. you are calculating a 2-ply. At 0-ply it is a contact position, but 
> depending on the move chosen it may become a race at 2-ply. Do you switch 
> neural nets within this evaluation?
> 

Any n-ply evaluation consists of a (large) number of neural net
evaluations (0-ply evaluation or static evaluation). The algorithm for a
static neural net evaluation is simply:

(a) determine type of position
(b) apply appropriate evaluator (one of 3 neural nets or bearoff database)

This means that an n-ply evaluation may be the average of equities
evaluated with different neural nets or databases. This is typical for
borderline positions, e.g., holding games where many rolls break
contact.

J?rn

pgp9DXr_3BmFO.pgp
Description: PGP signature