The Bethe Partition Function of Log-supermodular Graphical Models.
Nicholas RuozziPublished in: NIPS (2012)
Keyphrases
- partition function
- graphical models
- belief propagation
- approximate inference
- message passing
- free energy
- undirected graphical models
- probabilistic inference
- probabilistic model
- markov networks
- structure learning
- probabilistic graphical models
- exact inference
- loopy belief propagation
- bayesian networks
- random variables
- conditional random fields
- conditional independence
- factor graphs
- variational methods
- importance sampling
- energy minimization
- structured prediction
- exponential family
- belief networks
- random fields
- marginal probabilities
- latent variables
- closed form
- parameter estimation
- markov random field
- higher order
- upper bound
- prior knowledge
- support vector