Author: Alvaro Jose Povoa Cardoso
Date: 14:04:58 04/28/03
In reading "Parallel Retrograde Analysis on a Distributed system" by Henri Bal and Victor Allis, in the section where they explain the sequential algorithm they say: ..."For each non-end node, the value can be calculated once the values of all children are known, or earlier, if the node can obtain the highest value possible through one of its children. In the latter case, there is no need to determine the values of the other children." Then he goes on saying that we must associate 2 variables to each database entry: - UnknownChildren (the nº of children wich value is currentely unknown) - BestValue (The best value so far for the entry) Then he says: "Thus, if UnknownChildren has dropped to 0, BestValue is the final value of the position. Alternatively, once BestValue reaches a value which cannot be surpassed (e.g. value+1 in chess), then, regardless of the value of UnknownChildren, BestValue is the final value of the position" After all of the citations my question is: In the 2nd phase (the iterative propagation phase): If for example we have a position with say 10 children and currentely 9 children are unknown and one has a known (non-draw) value, how in the world do we know if that child is the best (or not) without knowing the rest of the (9)? Would that be by any chance connected to current number of passes we have done on the current database? Best regards, Alvaro Cardoso
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.