Multi-class SVMs: From Tighter Data-Dependent Generalization Bounds to Novel Algorithms

Keyphrases

• Class Size 33%
• Data Dependency 100%
• Duality Theory 33%
• Efficient Optimization 33%
• Existing Data 33%
• Fenchel Duality 33%
• Generalization Analysis 33%
• Generalization Bounds 100%
• Generalization Error Bound 33%
• Generalization Performance 33%
• Linear Dependence 33%
• Lp-norm Regularization 33%
• Multi-class Classification 66%
• Novel Algorithm 100%
• One-class Support Vector Machine (OCSVM) 100%
• Optimization Algorithm 33%
• Tight 100%

Computer Science

• Classification Algorithm 50%
• Generalization Error 50%
• Generalization Performance 50%
• Linear Dependence 50%
• Multi Class Classification 100%
• Multiclass Classification 100%
• Optimization Algorithm 50%
• Regularization 50%
• Support Vector Machine 100%

Engineering

• Classification Algorithm 50%
• Error Bound 50%
• Linear Dependence 50%
• Lp-Norm 50%
• Multiclass Classification 100%
• Regularization 50%
• Size Class 50%