MOKD: Cross-domain Finetuning for Few-shot Classification via Maximizing Optimized Kernel Dependence

Keyphrases

• Bilevel Optimization 33%
• Class Centroid 33%
• Class-specific 33%
• Cluster Structure 33%
• Cross-domain 100%
• Cross-domain Few-shot Classification 33%
• Data Representation 33%
• Few-shot Classification 100%
• Generalization Performance 33%
• High Similarity 33%
• Hilbert-Schmidt Independence Criterion 66%
• Labeled Data 33%
• Learning Representations 33%
• Meta-dataset 33%
• Metric Space 33%
• Nearest Centroid Classifier 100%
• Optimization Problem 33%
• Project Repository 33%
• Two-sample 33%
• Unseen Domain 33%

Computer Science

• Bilevel Optimisation 33%
• Cluster Structure 33%
• Data Representation 33%
• Generalization Performance 33%
• Meta Dataset 33%
• Metric Space 33%
• Nearest Centroid 100%
• Optimization Framework 33%
• Optimization Problem 33%