-
Notifications
You must be signed in to change notification settings - Fork 0
/
README
23 lines (18 loc) · 1.03 KB
/
README
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
Programs (in R) for the results in the paper:
Modifying Gibbs Sampling to Avoid Self Transitions
Radford M. Neal, 2024
The paper is in gibbs-self.pdf here, and at https://arxiv.org/abs/2403.18054
The algorithms described in the paper are implemented (in R) in the
file methods.r. Various checks for validity of these methods are
done in check-all.r.
The experimental runs reported in the paper can be reproduced with the
run-potts-8x8, run-potts-5x5, run-mix, run-beliefnet, and run-other
shell scripts. These runs take a long time - several days, in total.
The outputs of these runs are in the sub-directory 'runs', except for
the saved R workspaces, which would be quite large.
Checking out the small-runs branch of this repository will switch to
smaller runs (1/100 the length of the full runs), which take only
about a half hour or less for each shell script. Of course, the
asymptotic variance estimates are then much noisier. The results of
these smaller runs are in the 'runs' sub-directory when the small-runs
branch is checked out.