Supported Jobs and MethodsΒΆ

The following jobs and methods are available in FastPGM.

  • Structure learning: to learn the structure (a graph) of a BN

    • PC-Stable

  • Learning: to learn the structure (a graph) and parameters (conditional probability tables) of a BN

    • PC-Stable + maximum likelihood estimation

  • Exact inference: to infer the exact posterior distribution of unknown variables, given observations of some variables

    • Brute force

    • Junction tree

    • Variable elimination

  • Approximate inference: to infer the approximate posterior distribution of unknown variables, given observations of some variables

    • Loopy belief propagation

    • Probabilistic logic sampling

    • likelihood weighting

    • self-importance sampling (and variances)

    • AIS-BN

    • EPIS-BN

  • Classification: to categorize based on features (variables in BNs), accomplished through the building blocks of structure learning, parameter learning and inference.

  • Other functionalities related to BNs: such as sample generation, dataset and network format convertor, etc.

Please see knowledge base for the related basis and terminologies.