Spent today re-running simulation a number of times … but finally got there. Downloaded results, and then went through munging process.
Note to self: don’t see any reason why can’t run DirectoryMunger on a full dir of files — should still play ball.
March 31, 2009
March 30, 2009
Monday: stupidity strikes …
Looking at results from Grid, struggling to identify why the velocities never change. They never change, because the population never gets passed to the velocity calculation rules. Smooth, real smooth.
Idiot.
March 25, 2009
Wednesday: merged code base today …
Today made the changes to grid code that have been made to local code base. Attempted to re-run on the Grid, and appears to have run successfully. Now just need to sanity check the data, and then start to re-run the simulations.
March 24, 2009
Tuesday: fixed model logic
(NB direct linkage to previously mentioned activestate page: http://code.activestate.com/recipes/502240/ – this is the more up to date link)
Got the model running better now. Today implemented a different scaling rule, so that positions do not just progress linearly. This coupled with the work done on Friday, where the correct population is based to the flocking rules => better flocking behaviour.
Now just need to diff the changes I made and merge into the code used on the grid to re-run the experiments. (Note to self: need to learn how to wield Subversion to do branches / merges correctly!)
March 20, 2009
Friday: broken velocity scaling logic …
So I have the velocity rule:
What seems to happen is that if you exceed the max vel — let’s say its 10 — and come up with a vel of say 10.9416. Max vel / vel = a caling factor of 0.913946. Multiply this scaling factor by the vel of 10.9416 and you get …. 10. Queue permenant loop, and velocity set to 10. Marvelous. Hence you see birds just proceeding in straight lines across the environment.
How do other people do this? What happens if you ditch the velocity rules? Vergenet says to keep the vel bounding rules. How to implement then? The activestate page has an interesting section on velocity.mag ….
March 15, 2009
Sunday: data not quite legit it seems …
R command of the day:
par(mfrow=c(2, 2), cex=0.6, mar=c(4, 4, 1, 1)) plot(y, type="p")From http://www.stat.auckland.ac.nz/~paul/RGraphics/rgraphics.html — useful for graphing.
Appears data is screwy for some reason. More investigation ahoy!
March 13, 2009
Fri: stating it up …
Need to use factorial anova. See delivicous for details.
R snippets:
test = apply(tt2, MARGIN=2, FUN=mean, na.rm=TRUE) # calculate the mean for each column
t1 = read.table(“D:/Temp/Data/GridResults/Results/Stat1/output.txt”, sep = “,”, head=TRUE) # seems to be the preferred way to read in files
t2 = t1[,1:10] # how to lop off the trailing column if it is full of NA’s
tt2 = t(t2) # transpose data in the preferred format (it appears) for R
