Random Permutations for MCMC

Question

I'm new to JAGS (Just Another Gibbs Sampler) and I was wondering if it is possible to extend it to sample from a space of random permutations?

The reason I ask is that I came across this tutorial on extending JAGS, and one fo the required functions "logDensity" has a required input as a double however to work with permutations I would need to input a double.

http://www.cidlab.com/prints/wabersich2013extending.pdf

Also I don't have to work with just JAGS, if anyone has another suggestion that would work I would appreciate any input.

I could code the sampler myself, I realize this is a simple task however, I wanted to use dclone (R library that uses R with Jags) to make the process parallel. — steve3051980, Feb 26 '16 at 16:24

score 0 · Accepted Answer · edited May 23 '17 at 12:15

0

If the permutations are over 18 or fewer elements, you can represent a permutation as a double by computing its lexicographic rank. (It's possible to squeeze out a couple more elements by computing the rank as a 64-bit integer and then reinterpreting the bits as a double.)

edited May 23 '17 at 12:15

Community

1
1

answered Feb 26 '16 at 16:47

David Eisenstat

64,237
7
60
120

Thank you I was just also looking at the Lehmer code also. I think this will work. Thank you – steve3051980 Feb 28 '16 at 06:35

Random Permutations for MCMC

1 Answers1