Sparse Forward-Backward Using Minimum Divergence Beams for Fast Training Of Conditional Random Fields.
Chris PalCharles SuttonAndrew McCallumPublished in: ICASSP (5) (2006)
Keyphrases
- conditional random fields
- forward backward
- hidden markov models
- structured prediction
- supervised training
- restricted boltzmann machine
- sequence labeling
- probabilistic model
- maximum entropy
- error correcting output coding
- image labeling
- named entity recognition
- higher order
- pairwise
- graphical models
- markov random field
- information extraction
- crf model
- random fields
- web page prediction
- semi markov
- generative model
- structured learning
- training set
- conditional likelihood
- weighted sums
- protein fold recognition
- approximate inference
- parameter learning
- high dimensional
- image processing
- discriminative learning
- markov networks
- information retrieval
- natural language processing
- prior knowledge