Skip to content

Latest commit

 

History

History
105 lines (67 loc) · 3.46 KB

File metadata and controls

105 lines (67 loc) · 3.46 KB

Changelog

Breaking changes

  • Simplifies return types of bbn.FromFile function (#88)

Performance

  • Speed up factor operations by up to factor 2, through faster factor indexing (#90)
  • Optimize rearranging factors, with 10-20% speedup, also in methods using it (#92)

Other

  • Removes unused type ConflictingEvidenceError (#87)
  • Adds profiling for variable elimination (#89)
  • Adds benchmarks for the most crucial operations (#90, #92)
  • Adds more unit tests (#93)

Features

  • Makes sub-package ve for low-level variable elimination public (#73)
  • Solve utility nodes even when there are ignored policies, overwritten by "evidence" (#74, #75)
  • Adds a logic factor to combine bits from parents to a decimal integer (#83)

Documentation

  • Adds more documentation and code examples for library usage (#78, #79, #80, #81)
  • Adds a decision example for selecting a diagnostic test (#82)
  • Adds more logic example (#83, #84)

Features

  • Reworked library and apps to use Variable Elimination for inference (#42, #43, #44, #45, #46, #47, #48)
  • Implement multi-stage decision networks (#45, #60, #61)
  • Adds network training to the TUI app (#54)
  • Adds a toggle to the TUI app to ignore policies of decision nodes with evidence (#56)
  • Nodes can be moved in the TUI app with WASD keys (#57)
  • Save network from TUI app using Ctrl+S (#57)
  • Adds a help dialog to the TUI app (#58)
  • Optional total utility node for weighting individual utilities (#59)
  • Adds support for custom node colors in YAML files (#64)
  • Adds shortcuts for logic nodes like, and, or, if-then, etc. (#65)
  • Networks can have a longer info text, also shown in TUI app (#66)

Documentation

  • Adds oil drilling decision example (#48)
  • Adds robot decision example (#50)
  • Adds medical treatment decision example (#60)
  • Adds disease control decision example (#61)
  • Adds three examples for logic induction using well-known logical puzzles (#65, #66)
  • All examples have a detailed description that shows up in the TUI app (#66, #67)

Other

  • Release downloads now include all examples (#68)

Features

  • Uses likelihood-weighted sampling instead of simple rejection sampling (#39)

Documentation

  • Adds an earthquake sensor decision network example (#38, #40)

Other

  • Consistently check whether a new network is an acyclic graph (#40)

Features

  • Adds mouse support for setting outcomes and showing tables (#33)
  • Adds utility and decision nodes to form decision networks (#34)
  • Color nodes by type (nature, decision, utility) (#35)

Features

  • Support reading BIF-XML file format (#25)

Documentation

  • Adds "Dog Problem" example in BIF-XML format (#25)
  • Adds "Native Fish" example in YAML format (#28)
  • Adds "Mendel Genetics" example in YAML format (#29)
  • Extend Monty-Hall problem example by "Change Door" variable (#30)

Bugfixes

  • Fix outcome labels in probability table view (#27)

Initial release of BBN, the Bayesian Belief Network library and CLI/TUI tool for Go.