aomddBN
Compiles AND/OR Multi-Valued Decision Diagrams (AOMDDs) for Bayesian networks.
Author
Description
The library contains the implementations of the depth AND/OR search based compiler for AND/OR Multi-Valued Decision Diagrams (AOMDDs) for Bayesian Networks
Input format
The Bayesian network instance has to be specified in the Ergo file format.
For some example problems please see our Repository.
Usage
The algorithm is invoked with the following command line arguments:
aomddBN -f <input file>
-h <pseudo tree>
-a <algorithm>
-l <time limit>
-cb <cache bound>
-cp <constraint propagation>
-v <verbosity>Detailed step-by-step instructions are available here.
Parameters
Parameters for the algorithm are:
f: (string) specifies the input file name
h: (string) the heuristic to use for finding an variable elimination order by which to construct the And/Or search space:
hypergraph: i.e., hypergraph partitioning heuristic
minfill: i.e., min-fill elimination heuristic
a: (integer) specifies which algorithm to run. The following values can be used:
- 91: Depth-first AND/OR search with adaptive context based caching
l (integer) specifies the time limit in seconds. Default value is -1, which indicates no time limit
cb (interger) specifies the cache bound
cp (string): specifies the constraint propagation method. The following values can be used:
none: no constraint propagation is performed
bcp: constraint propagation via unit resolution over the CNF encoding of the zero probability CPT entries
sat: constraint propagation via full satisfiability over the CNF encoding of the zero probability CPT entries
v (string): verbosity. The following values can be used:
yes: enabled
no: disabled.
Example
aomddBN -f test.erg -h minfill -a 91 -l 100 -cb 10 -cp bcp -v yes
aomddBN will compile the Bayesian network specified in "test.erg" file by a depth-first AND/OR search with adaptive context-based caching guided by a minfill based pseudo tree. The time limit is set to 100 seconds, verbosity is enabled, the cache bound is set to 10, and will perform constraint propagation via unit resolution.
Download
A 32-bit Windows binary is available here.
If you are interested in obtaining the source code, please contact the program author.
References
[1] Robert Mateescu and Rina Dechter. Compiling Constraint Networks into AND/OR Multi-Valued Decision Diagrams (AOMDDs). In Proceedings of the International Conference on Principles and Practice of Constraint Programming, 2006. Link
[2] Robert Mateescu, Radu Marinescu and Rina Dechter. AND/OR Multi-Valued Decision Diagrams (AOMDDs) for Constraint Optimization. In Proceedings of the International Conference on Principles and Practice of Constraint Programming (CP), 2007. Link
[2] Robert Mateescu and Rina Dechter. AND/OR Multi-Valued Decision Diagrams (AOMDDs) for Weighted Graphical Models. In Proceedings of the International Conference on Uncertainty in Artificial Intelligence (UAI), 2007. Link