 |
 |
On Sat, 16 Apr 2005 20:39:58 +0200, =?iso-8859-1?B?SvZyZyBLcuRtZXI=?= <[log in to unmask]> wrote: It >works ok but we use it to evaluate fitness in a GA (roll eyes) which of >course multiplies any delays, so it might be worth restructuring. I just >found it hard to predict whether establishing a more complicated data >structure in the first place would save us any time - probably depends on >the small-world qualities of the individual graph. > Dear Joerg, This is just a quick reminder that your may get properties of small world networks with visibility graphs, but this is probably for trivial reasons. Indeed, isovists are computational artefacts, built in a way that a graph of the isovist points will, in general, contain points both close and far away (in Euclidean terms). Thus, if you build a graph out of inter-visible points, you'll probably get small world properties -trivial, isn't it? Now: 1. The small world phenomenon is a trick -it doesn't seem to mean much- and depends on the presence of long corridors so that the isovists contain points in the neighbourhood as well as faraway. So it is not a general phenomena (you’re not likely to find it, unless the plan contains these "corridors"); 2. Isovists haven't taken off in the Psychology/Neuroscience literature. So, be careful as the link between isovists and perception is yet to be established outside of the 300 or so members of this list. Most people would regard isovists as computationally interesting –but any theory relating them with human perception is “ad-hoc” at the time of writing. All the best, Rui >Joerg > > >-- >Joerg Kraemer, Jan-Oliver Kunze >Diplomanden FG Finn Geipel, Technische Universitaet Berlin >Ackerstrasse 71-76, Raum 438, 13355 Berlin, Germany +49 30 31472748 >[log in to unmask], [log in to unmask] > > > >> -----Original Message----- >> From: [log in to unmask] [mailto:[log in to unmask]] On >> Behalf Of Alasdair Turner >> Sent: Friday, April 15, 2005 10:06 PM >> To: [log in to unmask] >> Subject: Re: Seeking a list of space syntax algorithms >> >> Dear Jorg, >> >> In order to make graph connections, Depthmap simply uses an expanding >> ring around the current grid point -- I believe it's a fairly standard >> technique, although probably not the fastest -- if I were doing it again >> I would use a B*-tree (do a Google search to find online algorithms). I >> believe Yiorgos Chrysanthou (Google once more) coded the version used in >> Sanjay Rana's Isovist Analyst extension. >> >> As for integration the analysis, Depthmap just uses a straight forward >> breadth first search -- any algorithm book should be able to give you as >> good a algorithm as used in Depthmap -- there are no hidden >> optimisations, although I would say that it is very well written! :-) >> >> Alasdair >>