Closed GoogleCodeExporter closed 8 years ago
LISA / Getis-Ord make use of parallel threads for random trials. This greatly
complicates the use of a common seed. Perhaps we want to reconsider doing this
feature?
Original comment by mmcc...@gmail.com
on 21 Feb 2014 at 5:08
If I can find a pseudo random number generator, rng(val) with the property that
rng(seed+n) is a good random value (and fast) for successive values of n, then
could still use parallel processing with good/repeatable results using a common
starting seed. This will require some rewriting and testing to see that
results are good and algorithm speed isn't too slow.
Original comment by mmcc...@gmail.com
on 21 Feb 2014 at 5:16
Fully implemented for Getis-Ord and LISA. Feature will appear in GeoDa 1.5.27
or later. See attached images for new menu options and seed entry dialog.
Feature finished in SVN revision 3068
Original comment by mmcc...@gmail.com
on 25 Feb 2014 at 12:20
Attachments:
Original issue reported on code.google.com by
mmcc...@gmail.com
on 21 Feb 2014 at 2:54