OCI-SSL: Open Class-Imbalanced Semi-Supervised Learning With Contrastive Learning

Keyphrases

• Accuracy Improvement 25%
• Adverse Effects 25%
• Area under the Receiver Operating Characteristic (AUROC) 25%
• Bernoulli Sampling 25%
• CIFAR-10 25%
• Class Distribution 25%
• Class Imbalance 100%
• Consistency Regularization 25%
• Contrastive Learning 100%
• Feature Learning 25%
• Hard Negative Sampling 25%
• Imbalanced Class Distribution 25%
• In-distribution 25%
• Labeled Data 25%
• Learning Methods 25%
• Model Performance 25%
• Novel Class 25%
• One-Vs-All Classifier 25%
• Open Class 100%
• Out-of-distribution Data 25%
• Semi-supervised Learning 100%
• Semi-supervised Learning Algorithm 50%
• State-of-the-art Techniques 25%
• Supervised Contrastive Learning 25%
• Unknown Class 25%
• Unlabeled Data 100%

Computer Science

• Average Accuracy 16%
• Class Distribution 33%
• Class Imbalance 16%
• Contrastive Learning 100%
• Data Distribution 33%
• Learning Algorithm 33%
• Performance Model 16%
• Regularization 16%
• Representation Learning 16%
• Sample Distribution 16%
• Semisupervised Learning 100%
• Unlabeled Data 66%