Probabilistic Graphical Models

From BarikWiki
Jump to: navigation, search

Another Coursera class!

Probabilistic Graphical Models

Notes

  • Model is a declarative representation, that can be turned in different algorithms. Separate model from algorithm.
  • probability = lets us deal with uncertainty; established learning methods
  • graphical = complex systems through (graphs); joint distributions large, need to exploit structure
  • intuitive and compact data structure
  • directed and undirected, such as temporal and plate models
  • conditioning (conditional probability) / reduction
  • marginalization -> throw away
  • factor