Learning Maximum-A-Posteriori Perturbation Models for Structured Prediction in Polynomial Time.
Asish GhoshalJean HonorioPublished in: ICML (2018)
Keyphrases
- structured prediction
- markov networks
- maximum a posteriori
- maximum margin
- conditional random fields
- map estimation
- efficient learning
- image reconstruction
- em algorithm
- learning algorithm
- graphical models
- markov random field
- latent variables
- hidden variables
- probabilistic model
- bayesian framework
- structured prediction problems
- approximate inference
- belief networks
- convex optimization
- reinforcement learning
- bayesian inference
- parameter estimation
- active learning
- prior knowledge
- pattern languages
- training data
- feature selection