To Reverse the Gradient or Not: an Empirical Comparison of Adversarial and Multi-task Learning in Speech Recognition.
Yossi AdiNeil ZeghidourRonan CollobertNicolas UsunierVitaliy LiptchinskyGabriel SynnaevePublished in: ICASSP (2019)
Keyphrases
- speech recognition
- multi task learning
- multi task
- multiple tasks
- learning tasks
- gaussian processes
- hidden markov models
- language model
- speech recognizer
- pattern recognition
- speech recognition systems
- speech signal
- learning problems
- theoretical analysis
- high order
- speech synthesis
- transfer learning
- automatic speech recognition
- edge detection
- learning models
- machine learning
- inductive learning
- maximum likelihood
- supervised learning
- bayesian networks
- similarity measure
- learning algorithm