Variable Elimination

Probabilistic Graphical Model Exact Inferencing

Variable Elimination is an exact inference algorithm for Bayesian networks and Markov random fields.

Introduction

Implement Variable Elimination. Read Bayesian and Markov networks and evidence file specified using .uai format (http://graphmod.ics.uci.edu/uai08/FileFormat). Variable elimination algorithm has three steps:

1. Instantiate Evidence (Reduce the CPTs or factors)

2. Order the variables

3. Eliminate variables one by one along the order. To eliminate a variable, we compute a product of all functions that mention the variable and sum-out the variable from the result. Then we replace all functions that mention the variable with this new function

How to run

1. Navigate into the VariableElimination folder

2. Launch command prmpt

3. Run VE_exe.exe on command promt using following command

   $ java -jar VE_exe.jar

Sample Input and Output

$ java -jar VE_exe.jar BN_4.uai BN_4.uai.evid

---------------------------RESULT------------------------------

Maximum width: 19

Log Base e probability of evidence = -3.991165429944898

Log Base 10 probability of evidence = -1.7330288449608764

Probability of evidence = 0.01847816655753799

Total execution time = 17 seconds!

Output file: output1611.txt written successfully!

---

NOTE:

1. OutputXXXX.txt file is created with the above output in the same folder

2. BN_6.uai and BN_8.uai produces Out Of Memory Error: Java heap space because min order generated produces higher treewidth.