home *** CD-ROM | disk | FTP | other *** search
- Comments: Gated by NETNEWS@AUVM.AMERICAN.EDU
- Path: sparky!uunet!cs.utexas.edu!sdd.hp.com!swrinde!gatech!darwin.sura.net!paladin.american.edu!auvm!MATGEN.GE.CNR.IT!FRANKLIN
- X-Envelope-to: gis-l@UBVM.cc.buffalo.edu
- X-VMS-To: ICE622::IN%"gis-l@UBVM.cc.buffalo.edu"
- X-VMS-Cc: FRANKLIN
- Message-ID: <B0C007DDC0A00249@ICE.GE.CNR.IT>
- Newsgroups: comp.infosystems.gis
- Date: Tue, 22 Dec 1992 18:38:00 MET
- Sender: Geographic Information Systems Discussion List <GIS-L@UBVM.BITNET>
- From: Wm Randolph Franklin <FRANKLIN@MATGEN.GE.CNR.IT>
- Subject: Complexity of updating a Voronoi diagram
- Lines: 43
-
- On Dec 11, srobeson@bronze.ucs.indiana.edu (scott robeson),
- in an article on Re: Voronoi Tessalation GIS, said this:
-
- In some ill-conditioned distributions of control points,
- however, the Voronoi tesselation can change dramatically
- just by adding or removing a few points in strategic
- locations.
-
- The question is, how often does that occur in practice?
-
- If you're given all the points in advance, and process them
- in random order, then the expected total time is excellent,
- i.e., N*log(N) or N*log(N)^2 (I forget which). The
- expectation is over the random number generator in the
- program; every time you run the program it will take a
- slightly different time to produce the same output. There
- is no input data whose expected time is worse.
-
- This is an example of a Las Vegas randomized algorithm in
- computational geometry. It is a dramatic concept only about
- 5 years old, which leads to algorithms both faster and
- simpler. (The tradeoff is that they can be very hard to
- analyze.)
-
- If the user is specfying the points, and they must be
- processed "on-line" 1-by-1 as given, then the time can blow
- up to quadratic.
-
- I don't think that such a circumstance is worth worryinbg
- about since I'd doubt it would occur in a real example.
- However, even then, I could imagine a multi-level Voronoi
- diagram that might handle that. The idea would be to
- insert troublesome new points into a temporary, smaller data
- structure. From time to time merge the data structures.
- All queries must be done in the two data structures.
-
- Now let the "two" be a variable depending on the number of
- points.
-
- The idea was described by Jon Bentley and Mary Shaw (I
- think) as a general technique for dynamizing static data
- structures in IEEE Trans. on Software Engin. at least 15
- years ago.
-