A maximum entropy framework for semisupervised and active learning with unknown and label-scarce classes

Computer Science

• Maximum Entropy 100%
• Active Learning 100%
• Unlabeled Sample 50%
• Semisupervised Learning 50%
• Machine Learning 25%
• Learning Systems 25%
• Labeled Example 25%
• Sparse Solution 25%
• Feature Interaction 25%
• Based Active Learning 25%
• Common Multiple 25%

Keyphrases

• Unknown Categories 100%
• Overtraining 66%
• Learning Issues 33%
• UCI Machine Learning 33%
• Label Imbalance 33%
• Zero Weights 33%
• Active Learning System 33%
• Pool-based Active Learning 33%
• Hierarchical Classes 33%
• Irrelevant Features 33%
• Sparse Solution 33%