Learning Maximum-A-Posteriori Perturbation Models for Structured Prediction in Polynomial Time.
Asish GhoshalJean HonorioPublished in: CoRR (2018)
Keyphrases
- structured prediction
- markov networks
- maximum a posteriori
- maximum likelihood
- maximum margin
- conditional random fields
- efficient learning
- image reconstruction
- latent variables
- approximate inference
- graphical models
- expectation maximization
- convex optimization
- map estimation
- belief propagation
- prior knowledge
- learning algorithm
- markov random field
- probabilistic model
- learning tasks
- bayesian framework
- parameter estimation
- hidden markov models
- similarity measure
- image segmentation
- first order logic
- energy function
- em algorithm
- message passing
- bayesian inference
- hyperparameters
- machine learning
- structured prediction problems