Pairwise Decomposition of Directed Graphic Models for Performing Amortized Approximate Inference.
Peng LinChangsheng DouNannan GuZhiyuan ShiLili MaPublished in: ML4CS (1) (2022)
Keyphrases
- approximate inference
- pairwise
- belief propagation
- graphical models
- parameter estimation
- variational methods
- message passing
- probabilistic model
- factor graphs
- exact inference
- structured prediction
- markov random field
- conditional random fields
- markov chain monte carlo
- model selection
- higher order
- dynamic bayesian networks
- energy function
- bayesian networks
- loopy belief propagation
- machine learning
- variational approximation
- feature selection