Author: Bruce Cleaver
Date: 14:13:03 02/15/04
Go up one level in this thread
On February 15, 2004 at 14:21:25, Christophe Theron wrote: >On February 14, 2004 at 20:15:58, Bruce Cleaver wrote: > >>"Why not use a logarithmic scale based on the difference between the best >>possible move and the move under consideration?" >> >>Ron Rivest (he is the "R" in the RSA encryption algorithm) wrote a chess >>algorithm called min-max approximation, which computes the first derivative >>(really!!) of the score's change as a means to shape the search. It has >>somewhat the same flavor as your idea. >> >>It is really beautiful, but has two flaws: it is a best-first searcher >>(therefore exponential in memory), and heavily involves floating-point calcs. >>The first objection can be overcome in the standard way, but not the second. > > > >I see a third conceptual flaw: it's not intuitive. I don't see this idea as >trying to mimic a human chess player's thinking process. > >Even MTD(f) seems more intuitive than this. > >If I had to try new ideas, I would not go into that direction. > > > > Christophe Rivest explains that the idea is to expand the node whose change in evaluation (if any) will have the most effect upon the root node score. It *does* take a different frame of mind than alpha-beta or any of its variants.
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.