MPLP++: Fast, Parallel Dual Block-Coordinate Ascent for Dense Graphical Models.
Siddharth TouraniAlexander ShekhovtsovCarsten RotherBogdan SavchynskyyPublished in: ECCV (4) (2018)
Keyphrases
- graphical models
- coordinate ascent
- belief propagation
- image recovery
- probabilistic model
- random variables
- probabilistic inference
- probabilistic graphical models
- approximate inference
- structure learning
- bayesian networks
- markov networks
- map inference
- conditional independence
- single processor
- conditional random fields
- global convergence
- message passing
- factor graphs
- statistical relational learning
- depth map
- parallel processors
- parallel architectures
- image representation
- genetic algorithm