BayesNetInfer

This project implements Bayesian Network inference using three probabilistic inference algorithms:

1. Simple Inference (no optimizations).
2. Variable Elimination with ABC variable ordering.
3. Variable Elimination with heuristic-based variable ordering.

The project was developed as part of an academic assignment focusing on performance comparison between different inference methods.

Implemented in Java 1.8.

---

Input

The program receives two types of input files located in the same directory:

• input.txt:

  • First line: the name of the XML file that describes the Bayesian Network structure.
  • Following lines: queries to be evaluated.

• XML File (Network Description):
  A Bayesian network structured using XML format, including:

  • Variables (<VARIABLE> tags with <NAME> and <OUTCOME>).
  • CPTs (Conditional Probability Tables) inside <DEFINITION> tags.

Supported Queries

1. Joint Probability
   Format:

   P(E1=e1, E2=e2, ..., En=en)
   

   Calculates the joint probability for fully assigned variables.

2. Conditional Probability
   Format:

   P(Q=q|E1=e1, E2=e2, ..., Ek=ek), <algorithm_number>
   

   • Q: query variable.
   • E1..Ek: evidence variables.
   • <algorithm_number> indicates which algorithm to use: 1, 2, or 3.

If a query directly matches an existing CPT entry, it is answered immediately without computation.

---

Output

The program outputs the results to a file named output.txt, created in the same directory.
Each line corresponds to a query result in the following format:

<result_probability>,<number_of_additions>,<number_of_multiplications>

• Probability values are written with five decimal digits precision.
• In normalization steps, addition operations are counted but division operations are not counted as multiplications.
• If no addition/multiplication was needed, 0 is written in their place.

---

Algorithms Details

• Simple Inference (Algorithm 1):
  Direct calculation of probabilities without optimization.

• Variable Elimination with ABC Ordering (Algorithm 2):
  Variables eliminated based on a fixed alphabetical order (A → B → C...).

• Variable Elimination with Heuristic Ordering (Algorithm 3):
  Variables eliminated based on a custom heuristic to optimize efficiency.
  (Heuristic explanation must be provided in a separate details document.)

• Factor Multiplication Order:

  • Factors are multiplied in ascending order of size.
  • If sizes are equal, factors are ordered by the sum of ASCII values of their variables' names (smallest first).

Note: No Bayes-Ball or conditional independence checks are required.

---

License

This project is part of an academic assignment and is intended for educational use only.