HOW DOES UNLABELED DATA IMPROVE GENERALIZATION IN SELF-TRAINING? A ONE-HIDDEN-LAYER THEORETICAL ANALYSIS

Keyphrases

• Unlabeled Data 100%
• Hidden Layer 100%
• Self-training 100%
• Iterative Self-training 100%
• Training Paradigm 66%
• Neural Network 66%
• Shallow Neural Network 66%
• Convergence Rate 33%
• Deep Neural Network 33%
• Learning Improvement 33%
• Supervised Learning 33%
• Labeled Data 33%
• Performance Accuracy 33%
• Generalization Ability 33%
• Unlabeled Samples 33%
• Theoretical Characterization 33%
• Generalization Accuracy 33%
• Landscape Analysis 33%
• Semi-supervised Learning Algorithm 33%
• Training Convergence 33%
• Theoretical Understanding 33%
• Self-supervised Pre-training 33%

Computer Science

• Unlabeled Data 100%
• Self-Training 100%
• Neural Network 42%
• Convergence Rate 14%
• Deep Neural Network 14%
• Learning Algorithm 14%
• Supervised Learning 14%
• Semisupervised Learning 14%
• Generalization Ability 14%
• Layer Neural Network 14%
• Unlabeled Sample 14%
• Correctness 14%