Author: Alessandro Damiani
Date: 14:34:40 01/01/00
Go up one level in this thread
On December 31, 1999 at 16:39:08, Dan Andersson wrote: >I substituted ABC-Search for the usual AB search in MTDf, the straight undiluted >stuff. Look at MTDf, its AB with only one value instead of two. The reason I >switched to dynamic criterions was the fact that generating all legal moves near >the leaf node slows the search down, and the fact that I could get better >information about the need to extend search by minimaxing some select values in >my eval in the AB search I already have to do in IID. You can fix the problem w. >no fixed horizon by introducing a lower limit to the remaining depth, that >switches off the ABC search and goes to straight AB. > >Regards Dan Andersson How do you handle the transposition table? Since the search depth of a subtree is not known when it is searched (not a uniform subtree). A natural choice would be depth = max-depth + min-depth (max- and min-depth as described in the paper). Alessandro
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.