Submodular subset selection for large-scale speech training data.
Kai WeiYuzong LiuKatrin KirchhoffChris D. BartelsJeff A. BilmesPublished in: ICASSP (2014)
Keyphrases
- subset selection
- training data
- feature selection
- speech recognition
- test data
- small scale
- learning algorithm
- data sets
- hill climbing
- decision trees
- training set
- test set
- classification accuracy
- speech signal
- objective function
- labeled data
- prior knowledge
- automatic speech recognition
- supervised learning
- machine learning
- real world
- audio visual
- labelled data
- speaker identification
- broadcast news
- learned from training data
- generalization error
- classification models
- greedy algorithm
- tabu search
- training examples
- domain knowledge
- neural network