Mixup-breakdown: A Consistency Training Method for Improving Generalization of Speech Separation Models.
Max W. Y. LamJun WangDan SuDong YuPublished in: ICASSP (2020)
Keyphrases
- probabilistic model
- preprocessing
- prior knowledge
- significant improvement
- monte carlo simulation
- cost function
- detection method
- training process
- high accuracy
- em algorithm
- parameter estimation
- bayesian framework
- machine learning methods
- linear model
- machine learning
- linear regression
- statistical model
- error rate
- clustering method
- edge detection
- k means
- pairwise
- similarity measure