 |
 |
Hi Alasdair, You'll have to summarise ;) Rui > -----Original Message----- > From: [log in to unmask] [mailto:[log in to unmask]] On > Behalf Of Alasdair Turner > Sent: 20 December 2005 10:54 > To: [log in to unmask] > Subject: Re: [SPACESYNTAX] RA or RRA > > Indeed, Mike, and Lucas, and Sheep, and Bin (and everyone else > developing software) -- each of us has a responsibility to communicate > what are our software does. That's probably an academic duty, but I > think for most of us it is natural to try to do. > > However, I believe we need more than graph theoretic formulae to > communicate precisely what we mean: natual language, mathematical > formulae, and pseudo-code all have their place. One thing that > frustrates me is the limitation of mathematical formulae when it comes > to describe, for example, a procedural optimisation in the code. I also > feel that formulae often obfuscate simple procedures, and so I > personally often prefer pseudo-code. Of course, the only way someone > can really verify what the program actually does is to read the source > code, which returns to an argument about open source, but also raises > the problem of when we say the code does one thing, and in fact it does > not (e.g., a bug in Depthmap recently confused log base 10 and log base > e -- natural to do as in C++ 'log' means log base e, wheras > conventionally in mathematics, it means log base 10, and ln means log > base e -- had the code been open source, any one could have spotted > this, but the formula alone would not have sufficed, dutifully > transcribing the 'log'). > > As for 'the counting from 0', I started out programming in Fortran, and > although I prefer C++, I have nothing in particular for or against > 0-based indexing. However, the reasons for the radius difference do not > concern zero indexing, but whether you say a radius is measured in nodes > or steps. > > Alan shows the justification well with VGA, which is indeed how it > happened, but I happen to think this definition also fits logically into > axial analysis: I view radius as a 'distance' (albeit a step distance) > and therefore it's units to my mind should be steps. The same applies > in angular analysis, where I see radius as a sum of angles rather than a > number of nodes. > > Finally, in case anyone is concerned that this information is only > contained in a web discussion: the radius used in Depthmap is (I hope > clearly) explained in the Depthmap manual. > > Alasdair > > Michael Batty wrote: > > this conversation is almost making me look at my own software - I have > > always assumed that there are many formulas in space syntax that were > > not well formulated originally, and that these issues are resolved when > > programmed. For example the definition of integration in the Social > > Logic of Space is not in the reciprocal form that is needed when one > > speaks of high or low integration but this gets corrected in the > > software. Time for some standard graph theoretic statements I think of > > all the formulas used > > > > Mike > > > > At 13:37 19/12/2005, Lucas Figueiredo wrote: > > > >> Hello Sheep. > >> > >> No one is counting from 0. A single topological step means depth 1. > >> There is no topological step from a given line to itself. > >> > >> I know that arrays are confusing in some old programming languages (I > >> also programmed in ANSI C and Pascal) and sometimes we are obliged to > >> change outputs because such limitations of data structure. > >> > >> However, I read lots of papers using R3 and I always understood this > >> radius as 3 steps away, not 2. I also think that in my MSc > >> dissertation wrote that R3 (3 steps) is the local standard radius > >> (which is wrong). > >> > >> These things (interpretations, implementations) must be clear. > >> > >> That is why I reinforce that academic software must be properly > >> published AND cited - because it is part of the methodology you use in > >> experiments. I think we must put it in our software licence "cite it, > >> otherwise do not use it". > >> > >> Regards! > >> Lucas > >> > >> > >> > >> On 19/12/05, Nick Dalton <[log in to unmask]> wrote: > >> > Hi the code in axman was designed to be output compatible with the > >> > fortran code. as a 'C' programmer I found it an odd switch ( and > >> > still do). > >> > > >> > People typically count from one (1,2,3,4) it's only C programmers ( > >> > and their children) that count from 0. Fortran and Pascal (object > >> > pascal being the language of Axman) use 1 based indexs for arrays and > >> > so number systems. Everything also had to be compatible with the > >> > output of the social logic of space ( with the D value). > >> > > >> > Zero depth makes sense to me but non programmers got there first. > >> > > >> > so R2 = r3, r3=r4 and r2=r1, which can be said to eliminate a > >> problem ( no r1). > >> > > >> > Using R2 (ie old R3 )You may noticed more glitches in radius 2 due to > >> > poor micro-structure. > >> > > >> > sheep > >> > > >> > >Dear Lucas, > >> > > > >> > >Yes, this is the same in Depthmap: R2 is the equivalent of R3 in axman. > >> > > > >> > >I have always said that this makes sense: two steps away is to my > >> > >mind R2, not R3. > >> > > > >> > >As for handling low numbers of lines within (Depthmap's) radius two, > >> > >undefined values (nulls) are given in the (small number of) cases > >> > >where there are too few lines to calculate RRA. > >> > > > >> > >Alasdair > >> > > > >> > > > > > > > Michael Batty Director Centre for Advanced Spatial Analysis (CASA) > > University College London - 1-19 Torrington Place - London - WC1E 6BT UK > > _______________________________________________________________ > > > > tel 44 (0) 207 679 1781 fax 44 (0) 207 813 2843 mobile 44 (0) 7768 423 656 > > > > */_Personal Web Page_/* *.* http://www.casa.ucl.ac.uk/people/MikesPage.htm > > */_Recent Books_/* *.** /Cities and Complexity/ . * > > http://www.complexcity.info/ > > */GIS, Spatial Analysis and Modeling/* *.* http://www.casa.ucl.ac.uk/GIS/ > > > > > > > > > > -- > Alasdair Turner > Course Director > MSc Adaptive Architecture and Computation > Bartlett School of Graduate Studies > UCL Gower Street LONDON WC1E 6BT > > http://www.aac.bartlett.ucl.ac.uk/