bn-pp

BN++ Data Structures and Algorithms in C++ for (discrete) Bayesian networks and Markov networks.

Compilation

$ make clean && make

Usage

Bayes nets

$ ./bn -h
usage: ./bn /path/to/model.uai [/path/to/evidence.uai.evid TASK] [OPTIONS]
TASK:
-pr     solve partition task
-mar solve marginals task
OPTIONS:
-ls   compute partition using logical sampling
-lw   compute partition using (bounded-variance) likelihood weighting
-gs   compute partition using gibbs sampling
-sp   compute marginals using sum-product in factor graphs
-ve   compute inference using variable elimination
-mf   variable elimination using min-fill heuristic
-wmf  variable elimination using weighted min-fill heuristic
-md   variable elimination using min-degree heuristic
-bb   variable elimination using bayes-ball
-h    display help information
-v    verbose

To inspect the markov assumptions of asia model

$ ./bn ../models/asia.uai -v <../models/asia.markov.query >markov.result.txt

To check the local markov independencies of asia model

$ ./bn ../models/asia.uai <../models/asia.ind

To check variable dependencies of asia model

$ ./bn ../models/asia.uai <../models/asia.not.ind

To execute queries or check (in)dependencies on the prompt

$ ./bn ../models/asia.uai -ve -bb
>> Query prompt:
? query 1 | 0, 2
P(1|0,2) =
Factor(width:2, size:4, partition:2)
1 0
0 0 : 0.99000
0 1 : 0.95000
1 0 : 0.01000
1 1 : 0.05000
>> Executed in 0.04897ms.
? query 2, 3, 4
P(2,3,4) =
Factor(width:3, size:8, partition:1.00000)
2 3 4
0 0 0 : 0.34650
0 0 1 : 0.14850
0 1 0 : 0.00350
0 1 1 : 0.00150
1 0 0 : 0.18000
1 0 1 : 0.27000
1 1 0 : 0.02000
1 1 1 : 0.03000
>> Executed in 0.05496ms.
? ind 7,1|4,5
true
? ind 4,1|6
false
? quit

Markov nets

$ ./mn
usage: ./mn /path/to/model.uai /path/to/evidence.evid [OPTIONS]
OPTIONS:
-h    display help information
-v    verbose

To compute the partition function of a Markov network given evidence

$ ./mn ../models/grid3x3.uai ../models/grid3x3-PR.uai.evid
>> Query prompt:
? PR
partition = 14.8899
>> Executed in 1.3077ms.

To compute the all the marginals of a Markov network given evidence

$ ./mn ../models/grid3x3.uai ../models/grid3x3-MAR.uai.evid
>> Query prompt:
? MAR
>> Marginals:
Factor(width:0, size:1, partition:1)
: 1.00000
Factor(width:1, size:2, partition:1.00000)
1
0 : 0.00011
1 : 0.99989
Factor(width:1, size:2, partition:1.00000)
2
0 : 0.17139
1 : 0.82861
Factor(width:0, size:1, partition:1.00000)
: 1.00000
Factor(width:0, size:1, partition:1.00000)
: 1.00000
Factor(width:1, size:2, partition:1.00000)
5
0 : 0.42988
1 : 0.57012
Factor(width:1, size:2, partition:1.00000)
6
0 : 0.99122
1 : 0.00878
Factor(width:1, size:2, partition:1.00000)
7
0 : 0.99995
1 : 0.00005
Factor(width:1, size:2, partition:1.00000)
8
0 : 0.57011
1 : 0.42989
>> Executed in 2.73007ms.

Input

The input format is the uai model specification for BAYES and MARKOV networks UAI 2014 Inference Competition.