FastPGM: Fast Probabilistic Graphical Model Learning and Inference

FastPGM is an open-source C++ library that aims to help practitioners easily and efficiently apply probabilistic graphical models (PGMs), especially Bayesian network (BN) models to solve real-world problems. FastPGM exploits multi-core CPUs to achieve high efficiency. Key features of FastPGM are as follows:

• Wide coverage of different tasks and algorithms related to PGMs, including structure learning, parameter learning, exact inference and approximate inference.

• Support classification, through the building blocks of structure learning, parameter learning and inference.

• Support Python interfaces.

• Support PGM sample generation, dataset and network format convertor, etc.

Getting Started

• Installation
  • Prerequisites
  • Download
  • Build
    • For Linux Users:
    • For MacOS (Apple Silicon) Users:
    • For MacOS (Intel) Users:
• First Example
• Supported Jobs and Methods
• Citing FastPGM
• Dependency

Next Steps

• FastPGM Parameters
  • Global Parameters
  • Parameters Related to Structure Learning
  • Parameters Related to Parameter Learning
  • Parameters Related to Approximate Inference
  • Parameters Related to Other Functionalities
  • Parameters for Files
• Python Interfaces
  • Prerequisites
  • Installation
  • Running FastPGM
  • Parameters

Knowledge Base

• Basics of FastPGM
  • Probabilistic Graphical Model
    • Bayesian Network
    • Markov Random Field
  • Structure Learning
    • Score-based Methods
    • Constraint-based Methods
  • Inference
• Frequently Asked Questions
  • Q1
  • Q2
  • Q3