Max-Margin Transducer Loss: Improving Sequence-Discriminative Training Using a Large-Margin Learning Strategy.
Rupak Vignesh SwaminathanGrant P. StrimelAriya RastrowSri Harish MallidiKai ZhenHieu Duy NguyenNathan SusanjAthanasios MouchtarisPublished in: ICASSP (2024)
Keyphrases
- learning strategies
- discriminative training
- log linear
- max margin
- log linear models
- hidden markov models
- maximum likelihood
- probabilistic interpretation
- support vector machine
- markov networks
- posterior probability
- maximum margin
- conditional random fields
- online learning
- structured prediction
- active learning
- multi class
- latent variables
- automatic speech recognition
- multi label classification
- efficient learning
- training data
- classification algorithm
- co training
- data sets
- multi label
- structured learning
- pairwise