Skip to content

BayesCalc2

A powerful Bayesian network calculator with interactive REPL

BayesCalc2 is a command-line tool for defining, analyzing, and querying Bayesian networks. It provides an intuitive syntax for specifying probabilistic relationships and a rich set of commands for probability inference.

Features

  • 🔢 Bayesian Network Definition: Simple, human-readable .net file format
  • 🧮 Probability Queries: Natural syntax like P(Rain|GrassWet)
  • 📊 Network Visualization: Generate PDF diagrams of your networks
  • 🔍 Advanced Inference: Variable elimination algorithm for efficient computation
  • 💻 Interactive REPL: Tab completion, command history, and helpful feedback
  • 📝 Batch Processing: Run multiple commands from script files
  • 🔗 Boolean Shorthand: Concise syntax for binary variables (Rain vs Rain=True)

Quick Start

Installation

pip install bayescalc2

Or install from source:

git clone https://github.com/ljp/bayescalc2.git
cd bayescalc2
pip install -e .

Your First Network

Create a simple network file rain.net:

variable Rain {True, False}
variable Sprinkler {On, Off}
variable GrassWet {Yes, No}

Rain { P(True) = 0.2 }
Sprinkler { P(On) = 0.1 }

GrassWet | Rain, Sprinkler {
    P(Yes | True, On) = 0.99
    P(Yes | True, Off) = 0.90
    P(Yes | False, On) = 0.85
    P(Yes | False, Off) = 0.05
}

Run Interactive Mode

bayescalc rain.net

Then query probabilities:

BayesCalc> P(Rain|GrassWet=Yes)
P(Rain=True | GrassWet=Yes) = 0.6203

Batch Mode

Create a command file queries.txt:

P(Rain)
P(GrassWet=Yes|Rain=True)
visualize output.pdf

Run it:

bayescalc rain.net -b queries.txt

Documentation

Example Networks

BayesCalc2 includes several example networks:

  • Medical Diagnosis: Disease testing with false positives/negatives
  • Weather Prediction: Rain, sprinkler, and grass wetness
  • Student Performance: Exam results based on intelligence and difficulty
  • Asia Chest Clinic: Medical diagnosis network from literature
  • Monty Hall Problem: Classic probability puzzle

See the Examples page for detailed explanations.

Architecture

BayesCalc2 uses a pipeline architecture:

.net files → Lexer → Parser → BayesianNetwork → Inference Engine
                                        ↓
                            QueryParser ← User Queries

Key components:

  • Lexer: Tokenizes .net files
  • Parser: Builds abstract syntax tree and network model
  • BayesianNetwork: Data model for variables, domains, and CPTs
  • Inference: Variable elimination algorithm for probability computation
  • REPL: Interactive shell with tab completion

Contributing

Contributions are welcome! Please see the Developer Guide for:

  • Setting up the development environment
  • Running tests
  • Code style guidelines
  • Architecture overview

License

BayesCalc2 is released under the MIT License. See LICENSE for details.