An improved Belief Propagation algorithm finds many Bethe states in the random field Ising model on random graphs.
Gabriele PeruginiFederico Ricci-TersenghiPublished in: CoRR (2017)
Keyphrases
- belief propagation
- markov random field
- factor graphs
- random fields
- probabilistic model
- graphical models
- parameter estimation
- energy function
- message passing
- objective function
- loopy belief propagation
- random graphs
- expectation maximization
- partition function
- bayesian framework
- k means
- graph cuts
- em algorithm
- dynamic programming
- approximate inference
- np hard
- conditional random fields
- non stationary
- maximum a posteriori
- pairwise
- generalized belief propagation
- image matching
- search space
- higher order
- free energy
- maximum likelihood
- power law
- markov networks
- constraint satisfaction problems
- bayesian inference
- image restoration
- phase transition
- maximum entropy
- energy minimization
- latent variables
- reinforcement learning