Unsupervised re-scoring of observation probability in viterbi based on reinforcement learning by using confidence measure and HMM neighborhood.
Carlos MolinaNéstor Becerra YomaFernando HuenupánClaudio GarretónPublished in: INTERSPEECH (2007)
Keyphrases
- confidence measure
- hidden markov models
- reinforcement learning
- viterbi algorithm
- confidence values
- sequential data
- supervised learning
- confidence measures
- hidden state
- speech recognition
- function approximation
- probability distribution
- unsupervised learning
- multi stream
- reinforcement learning algorithms
- markov decision process
- conditional random fields
- data driven
- semi supervised
- state space
- machine learning
- highly accurate
- gesture recognition
- forward backward
- handwritten word recognition
- temporal difference
- scoring model
- model free
- markov model
- optimal control
- conditional probabilities
- image matching
- transfer learning
- optimal policy
- active learning
- multi agent
- high quality