Research map / Continual learning / Class-incremental
Semantic Drift Compensation for Class-Incremental Learning
Lu Yu, Bartlomiej Twardowski, Xialei Liu, Luis Herranz, Kai Wang, Yongmei Cheng, Shangling Jui, Joost van de Weijer
Check your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use embedding-network fine-tuning with metric learning instead of softmax classification | Class-incremental learning · CUB-200-2011 | average incremental accuracy: +23.9% | Improved |
| Use embedding-network fine-tuning with metric learning instead of softmax classification | Class-incremental learning · CIFAR100 | average incremental accuracy: +32.6% | Improved |
| Use embedding-network fine-tuning with metric learning instead of softmax classification | Class-incremental learning · ImageNet-Subset | average incremental accuracy: +42.9% | Improved |
| Add SDC to E-FT (ablation) | E-FT · CIFAR100, T11 | average incremental accuracy: 37.4 -> 44.8 | Improved |
| Add SDC to E-FT (ablation) | E-FT · ImageNet-Subset, T11 | average incremental accuracy: 47.4 -> 50.9 | Improved |
| Add SDC to E-LwF (ablation) | E-LwF · CIFAR100, T11 | average incremental accuracy: 46.1 -> 46.1 | No clear effect |
| Add SDC to E-LwF (ablation) | E-LwF · ImageNet-Subset, T11 | average incremental accuracy: 51.5 -> 50.5 | Hurt |
| Add SDC to E-EWC (ablation) | E-EWC · CIFAR100, T11 | average incremental accuracy: 40.8 -> 46.1 | Improved |
| Add SDC to E-EWC (ablation) | E-EWC · ImageNet-Subset, T11 | average incremental accuracy: 49.5 -> 51.5 | Improved |
| Add SDC to E-MAS (ablation) | E-MAS · CIFAR100, T11 | average incremental accuracy: 43.1 -> 46.3 | Improved |
| Add SDC to E-MAS (ablation) | E-MAS · ImageNet-Subset, T11 | average incremental accuracy: 50.8 -> 51.2 | Improved |
| Replace triplet loss with Multi-similarity loss (ablation) | E-FT · CUB-200-2011, T6 | average incremental accuracy: +13.3% | Improved |
Closest papers by meaning
- Class-Incremental Learning: Survey and Performance Evaluation on Image Classification (TPAMI 2022 · similarity 0.91)
- Incremental Embedding Learning via Zero-Shot Translation (AAAI 2021 · similarity 0.90)
- A Model or 603 Exemplars: Towards Memory-Efficient Class-Incremental Learning (ICLR 2023 · similarity 0.90)
- Navigating Semantic Drift in Task-Agnostic Class-Incremental Learning (ICML 2025 · similarity 0.89)
- Class-Incremental Learning by Knowledge Distillation With Adaptive Feature Consolidation (CVPR 2022 · similarity 0.89)
- Incremental Learning via Rate Reduction (CVPR 2021 · similarity 0.89)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
