Training data reduction in deep neural networks with partial mutual information based feature selection and correlation matching based active learning.
Jian ZhengWei YangXiaohua LiPublished in: ICASSP (2017)
Keyphrases
- active learning
- neural network
- feature selection
- training data
- mutual information
- training examples
- training process
- learning algorithm
- training set
- supervised learning
- annotation effort
- classification accuracy
- imbalanced data classification
- unlabeled data
- selective sampling
- classification models
- machine learning
- image registration
- sample selection
- labeled data
- matching algorithm
- generalization error
- back propagation
- support vector machine
- decision trees
- random sampling
- text categorization
- neural network model
- test data
- pattern recognition
- test set
- pattern matching
- experimental design
- matching process
- genetic algorithm
- naive bayes
- semi supervised learning
- semi supervised
- feature reduction
- batch mode
- fuzzy logic
- high dimensionality
- dimensionality reduction
- feature set
- information gain
- data sets
- labeling effort
- support vector
- reduction method
- text classification
- model selection
- feature subset
- reinforcement learning
- relevance feedback
- feature extraction
- data reduction
- transfer learning
- feature points
- class labels
- knn
- learning strategies
- prior knowledge
- training samples