Newsgroups: comp.ai.games
Path: cantaloupe.srv.cs.cmu.edu!nntp.club.cc.cmu.edu!goldenapple.srv.cs.cmu.edu!das-news2.harvard.edu!oitnews.harvard.edu!news.dfci.harvard.edu!camelot.ccs.neu.edu!news.mathworks.com!news.maxwell.syr.edu!news.bc.net!torn!kwon!watserv3.uwaterloo.ca!undergrad.math.uwaterloo.ca!crpalmer
From: crpalmer@wheat.uwaterloo.ca (Chris Palmer)
Subject: Re: Dykstra v. A* (Was Re: Prim's v.s A*)
Sender: news@undergrad.math.uwaterloo.ca (news spool owner)
Message-ID: <E7rqCw.nyH@undergrad.math.uwaterloo.ca>
Date: Fri, 28 Mar 1997 19:34:08 GMT
References: <33268E08.315CDBAD@hig.no> <5h5j8n$6sc@scam.XCF.Berkeley.EDU> <AMITP.97Mar27144311@Ghoti.Stanford.EDU> <5hf0l3$o1l@scam.xcf.berkeley.edu>
Nntp-Posting-Host: wheat.uwaterloo.ca
Organization: MultiText Project
Lines: 28

In article <5hf0l3$o1l@scam.xcf.berkeley.edu>,
Jason Alan Nordwick <nordwick@scam.XCF.Berkeley.EDU> wrote:
>In article <AMITP.97Mar27144311@Ghoti.Stanford.EDU>,
>Amit J Patel <amitp@Ghoti.Stanford.EDU> wrote:
>"In general, A* "should" have an advantage because it has extra
>"knowledge of the graph, while the usual graph algorithms do not.  For
>"example, if you are trying to travel from London to Paris, A* will
>"tend to spend more time looking to the south, while most graph
>"algorithms look equally in all directions.
>
>However, it also has a larger overhead associated with it, and it
>horrible when the graphs are strongly connected, b/c it is a good
>tree-walking algorithm, but chunks on graphs.

It's trivial to add the same heuristic function to Djikstra's algorithm
to get a goal directed search.  I have yet to try an in depth comparison
between the two searching techniques but I'd expect them to explore
almost identical areas of the graph given the same function.

The real advantage in the A* algorithm is the fact that it works on
very large search domains.  For Djikstra's algorithm, you require a
table with size proportional to the graph you are searching.  

Cheers,
Chris.
-- 
Mail: crpalmer@undergrad.uwaterloo.ca
Homepage: http://www.undergrad.math.uwaterloo.ca/~crpalmer/
