Code for MAP inference in graphical models
8/8/2012
This code is written by Amir Globerson, David Sontag, Do Kook Choe, & Yitao Li,
and implements the algorithms described in the following three papers:
Efficiently Searching for Frustrated Cycles in MAP Inference.
David Sontag, Do Kook Choe, and Yitao Li
Uncertainty in Artificial Intelligence (UAI) 28. Catalina Island, United States. 2012.
Tightening LP Relaxations for MAP using Message Passing
David Sontag, Talya Meltzer, Amir Globerson, Tommi Jaakkola and Yair Weiss
Uncertainty in Artificial Intelligence (UAI). Helsinki, Finland. 2008.
Fixing max-product: Convergent message passing algorithms for MAP LP-relaxations
Amir Globerson, Tommi Jaakkola
Advances in Neural Information Processing Systems (NIPS) 21. Vancouver, Canada. 2007.
MPLP is free software; you can redistribute it and/or modify it
under the terms of the GNU General Public License as published by
the Free Software Foundation
MPLP is distributed in the hope that it will be useful, but
WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
General Public License for more details.
You should have received a copy of the GNU General Public License
along with MPLP; see the file gpl.txt. If not, write to the Free
Software Foundation, 59 Temple Place - Suite 330, Boston, MA
02111-1307, USA.
Usage
1) Run 'make'. This will compile the C++ code and create the binary 'solver'.
2) Run the solver as follows:
./solver grid4x4.uai grid4x4.uai.evid 234 MPE grid4x4.log
or simply
./solver grid4x4.uai
3) The MAP assignment will be the *last* line of the ".MPE" file that is created.
More details:
- 'grid4x4.uai' is the input factor graph in UAI format, described here:
http://www.cs.huji.ac.il/project/PASCAL/fileFormat.php
We support any factor graph on discrete variables, not just pairwise MRFs.
- 'grid4x4.uai.evid' is the evidence file, also described above (we allow at
most one line of evidence). Default is 'no.evid', which is a file with a single '0' denoting that there are zero lines of evidence.
- '234' is the random seed. Default is '0'.
- 'MPE' specifies that this is doing MAP inference (no other option allowed)
- 'grid4x4.log' is an optional log file that is output, useful for creating
plots. The first column is cummulative time in seconds, the second column
is the dual objective value, and the third column is the value of the best
assignment found so far. We also output the number of triplet clusters
added and the amount of time used by the cycle search algorithm.
You can download several more example UAI files here:
http://www.cs.huji.ac.il/project/PASCAL/showExample.php
We made two extensions to the UAI file format:
- If input filename ends in UAI.LG, we expect log-potentials.
- If the number of entries for a factor table is listed as -1, we interpret
this factor to be a Potts potential and initialize it to be 0 if x_i = x_j,
and the value provided otherwise. The stereo vision files provide an example
(see below).
Data used in UAI 2012 paper