[Travelr-announce] TravelR Release 0.1-3

[Announcements related to the TravelR project]

A major update of *travelr *has been posted to the pre-alpha SCM 
section, with some new documentation, working ("at least it runs without 
crashing") traffic assignment including all the base algorithms (notably 
Frank-Wolfe and ParTan), revised iterative proportional fitting function 
(I never could figure out the purpose of R's "sweep" function -- thanks 
ODOT'ers!  Now ipf can do as many dimensions on your array as you 
like...), plus there is a pair of neat (at least, I think they're neat) 
new functions:

   1. an adaptive quadrature function for computing numerical integrals
      of cost functions, which you need to do equilibrium optimization
      objective functions -- it's always faster to code the integral
      analytically, but the numerical integration adds to the toolbox
   2. a function for re-aggregating matrices using an equivalency
      table.  So you can convert your census block data to TAZs, or add
      up the TAZs into districts, or sub-divide counties into TAZs,
      etc.  This one is still pretty primitive and may not work smoothly
      with some obvious things, such as applying it to two-dimensional
      margins of three-dimensional arrays in order to map each sub-table
      into the same district structure

There are some new tests and examples (many more to come).  Feel free to 
try something out and contribute the code back as a new or revised 
test:  even if (and ESPECIALLY if) the function crashes doing something 
obvious, or you can't get it to do what you want, send me code that 
reproduces the problem, I'll fix it (either the function or the 
documentation), and I'll incorporate the test in the standard sequence.  
I haven't formalized enough tests yet, since I'm impatient to get to the 
"feature complete" stage (it's very close), but more and more are 
suggesting themselves and making their way into the "to-do" list.  I do 
intend to ramp up R-Forge's "tracker" feature so users can easily make 
bug reports, feature requests, request support (not likely to be highly 
responsive), etc.

All my TravelR energy now is going into a couple of large extended 
example that will become a vignette (one a whole model, the other a 
tutorial on how to code up a new assignment algorithm), plus 
incrementally adding to and polishing the documentation and examples 
(plus the comments in the code), some of which are a bit thin.  On the 
other hand, each time I've started digging into that, I've had to 
overhaul the code in some way or another.  Stability is still a way off...

Happy modeling!
Jeremy "R"

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.r-forge.r-project.org/pipermail/travelr-announce/attachments/20100503/1b58e647/attachment.htm>