Mitigating Forgetting in Continual Learning via Contrasting Semantically Distinct Augmentations

Keyphrases

• Continual Learning 100%
• Contrastive Learning 100%
• Online Continual Learning 75%
• Catastrophic Forgetting 50%
• Superior Performance 25%
• Recent Advances 25%
• Computational Cost 25%
• Learning Mechanisms 25%
• Mini 25%
• Sampling Strategy 25%
• Learning Objectives 25%
• System Size 25%
• Model Design 25%
• Feature Representation 25%
• Learning Technologies 25%
• Label Information 25%
• CIFAR-100 25%
• Model Bias 25%
• Classification Mechanism 25%
• Model Stability 25%
• Learning Scenario 25%
• Art Performance 25%
• CIFAR-10 25%
• Softmax Classifier 25%
• Model Learning 25%
• Cross-entropy Loss 25%
• Generalized Features 25%
• Augmentation Techniques 25%
• ImageNet Dataset 25%
• Non-stationary Data Streams 25%
• Strong Augmentation 25%
• Class-incremental 25%
• Technology for Learning 25%

Computer Science

• Contrastive Learning 100%
• Data Stream 25%
• Computational Cost 25%
• Superior Performance 25%
• Learning Mechanism 25%
• Learning Technology 25%
• Art Performance 25%
• Sample Distribution 25%