Information-Theoretic Bounds on Transfer Generalization Gap Based on Jensen-Shannon Divergence

Keyphrases

• Information-theoretic Bounds 100%
• Jensen-Shannon Divergence 100%
• Generalization Gap 100%
• Data Distribution 75%
• Training Data 50%
• Source Distribution 50%
• Training Loss 50%
• Information Theory 25%
• Divergence 25%
• Transfer Students 25%
• Two-parameter 25%
• Transfer Learning 25%
• Numerical Examples 25%
• Mutual Information 25%
• Kullback-Leibler Divergence 25%
• Target Distribution 25%
• Testing Data 25%
• Learning Data 25%
• Population Loss 25%
• Risk Minimization 25%
• Cumulant Generating Function 25%
• Excess Risk 25%
• Unbounded Loss Function 25%
• New Upper Bound 25%
• Domain Shift 25%

Computer Science

• Data Distribution 100%
• Training Dataset 66%
• Source Distribution 66%
• Transfer Learning 33%
• Numerical Example 33%
• Mutual Information 33%
• Risk Minimization 33%
• Leibler Divergence 33%
• Weighted Average 33%
• Target Distribution 33%
• Population Loss 33%
• Testing Data Set 33%
• Parameter Family 33%

Mathematics

• Jensen-Shannon Divergence 100%
• Data Distribution 75%
• Upper Bound 50%
• Minimizes 25%
• Numerical Example 25%
• Kullback-Leibler Divergence 25%
• Loss Function 25%
• Generating Function 25%
• Mutual Information 25%
• Parameter Family 25%
• Transfer Learning 25%
• Training Dataset 25%
• Weighted Average 25%
• Excess Risk 25%
• target distribution π 25%
• Cumulants 25%