Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Search tree vs. recursion

Author: Dan Andersson

Date: 12:28:52 11/27/01

Go up one level in this thread


I think you are confusing things. The depth first search (recursive search)
takes one extra memory slot per extra ply searched. The hash is only there to
improve the efficiency of the search.The Search tree will grow at the rate of
branching per extra ply. That is 40 times more memory per ply. Unless you cut or
reduce the the tree you will be overwhelmed. The space of a memory slot will
also expand with 5< bits per full ply. Due to the need for longer pointers. That
in itself makes one realize that the search tree in memory, needs a lot of
optimizations. To overcome the space complexity of storing the search tree. You
will incur some other overhead like: Sorting nodes, Deleting nodes and walking
pointers up and down. There exist some good best first algorithms that you might
look at. And the best IMO is SSS. But Aske Plaat showed that it was possible to
reformulate SSS in a depth first variation. And then went to show that there
existed a much better variant of that depth first search.
Links:
http://www.xs4all.nl/~aske/
http://www.xs4all.nl/~aske/pubs.html look for Research Re: search & Re-search
also read up on Conspiracy Number Search and Proof Number Search


MvH Dan Andersson



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.