Research map / Continual learning / Class-incremental
Prototype Augmentation and Self-Supervision for Incremental Learning
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PASS with prototype augmentation and self-supervision instead of the best non-exemplar baseline | CIFAR-100 class-incremental learning · CIFAR-100, 10 phases | accuracy: 29.3% | Helped |
| Use PASS with prototype augmentation and self-supervision instead of the best non-exemplar baseline | TinyImageNet class-incremental learning · TinyImageNet, 10 phases | accuracy: 25.2% | Helped |
| Add SSL label augmentation | CIFAR-10 incremental classification · CIFAR-10, novel classes | average incremental task accuracy: 18.81 -> 50.50 | Helped |
| Add SSL label augmentation | CIFAR-100 incremental classification · CIFAR-100, novel classes | average incremental task accuracy: 30.54 -> 41.46 | Helped |
| Add SSL to KD without protoAug (ablation) | CIFAR-100 class-incremental learning · CIFAR-100, 10 phases | accuracy: 6.04 -> 8.46 | Helped |
| Add SSL to KD without protoAug (ablation) | TinyImageNet class-incremental learning · TinyImageNet, 10 phases | accuracy: 4.70 -> 6.53 | Helped |
| Add protoAug to KD without SSL (ablation) | CIFAR-100 class-incremental learning · CIFAR-100, 10 phases | accuracy: 6.04 -> 39.80 | Helped |
| Add protoAug to KD without SSL (ablation) | TinyImageNet class-incremental learning · TinyImageNet, 10 phases | accuracy: 4.70 -> 26.52 | Helped |
| Add SSL to KD+protoAug (ablation) | CIFAR-100 class-incremental learning · CIFAR-100, 10 phases | accuracy: 39.80 -> 49.03 | Helped |
| Add SSL to KD+protoAug (ablation) | TinyImageNet class-incremental learning · TinyImageNet, 10 phases | accuracy: 26.52 -> 39.28 | Helped |
| Remove KD from the full PASS method (ablation) | PASS class-incremental learning · CIFAR-100 and TinyImageNet | accuracy | Hurt |
Closest papers by meaning
- eTag: Class-Incremental Learning via Embedding Distillation and Task-Oriented Generation (AAAI 2024 · similarity 0.88)
- Non-exemplar Online Class-Incremental Continual Learning via Dual-Prototype Self-Augment and Refinement (AAAI 2024 · similarity 0.88)
- Long-Tail Class Incremental Learning via Independent Sub-prototype Construction (CVPR 2024 · similarity 0.87)
- Multi-layer Rehearsal Feature Augmentation for Class-Incremental Learning (ICML 2024 · similarity 0.87)
- A Model or 603 Exemplars: Towards Memory-Efficient Class-Incremental Learning (ICLR 2023 · similarity 0.87)
- Autoencoder-Based Hybrid Replay for Class-Incremental Learning (ICML 2025 · similarity 0.86)
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.