On the Importance of Firth Bias Reduction in Few-Shot Classification

Keyphrases

• Bias Reduction 100%
• Few-shot Classification 100%
• Cosine 33%
• Reduction Method 33%
• Multinomial Logistic 33%
• Maximum Likelihood Estimator 16%
• Easy-to-implement 16%
• Computer Vision 16%
• Sampling numbers 16%
• First Order Term 16%
• Challenging Tasks 16%
• Class number 16%
• Feature Representation 16%
• KL Divergence 16%
• Optimization Objectives 16%
• Statistical Machine Learning 16%
• Small Sample Bias 16%
• Few-shot 16%
• Class Assignment 16%
• Label Distribution 16%
• Assignment Probabilities 16%
• Imbalanced Data 16%
• Cross-entropy Loss 16%
• Logistic Classification 16%
• Cross-domain Data 16%

Computer Science

• Image Classification 100%
• Machine Learning 50%
• Computer Vision 50%
• maximum-likelihood 50%
• Optimisation Objective 50%
• Class Assignment 50%
• Label Distribution 50%
• Imbalanced Data 50%

Mathematics

• Bias Reduction 100%
• Cosine 33%
• Maximum Likelihood Estimator 16%
• Smaller Sample 16%
• Probability Assignment 16%
• Cross-Entropy 16%