Research map / Continual learning / Class-incremental
PASS++: A Dual Bias Reduction Framework for Non-Exemplar Class-Incremental Learning
Fei Zhu, Xu-Yao Zhang, Zhen Cheng, Cheng‐Lin Liu
Read the paperCheck 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 |
|---|---|---|---|
| add protoAug to KD-only baseline (ablation) | PASS++ · CIFAR-100 (T=10) | last accuracy (%): 8.46 -> 39.80 | Improved |
| add self-supervised transformation (SST) to protoAug (ablation) | PASS++ · CIFAR-100 (T=10) | last accuracy (%): 39.80 -> 49.03 | Improved |
| add hardness-aware protoAug (ablation) | PASS++ · CIFAR-100 (T=10) | last accuracy (%): 49.03 -> 55.26 | Improved |
| add multi-view class-node ensemble (ablation) | PASS++ · CIFAR-100 (T=10) | last accuracy (%): 55.26 -> 57.69 | Improved |
| add hardness-aware protoAug (ablation) | PASS++ · CIFAR-100 (T=10) | forgetting (%): 30.25 -> 21.94 | Improved |
| add multi-view ensemble (ablation) | PASS++ · CIFAR-100 (T=10) | forgetting (%): 21.94 -> 23.00 | Hurt |
| add SST to baseline in zero-cost CIL (ablation) | PASS++ · CIFAR-100 zero-cost CIL (NCM) | accuracy on all classes (%): 45.16 -> 54.42 | Improved |
| add SST to baseline in zero-cost CIL (ablation) | PASS++ · CIFAR-10 zero-cost CIL (NCM) | accuracy on all classes (%): 49.62 -> 58.74 | Improved |
| use multi-view class-node ensemble instead of vanilla rotation ensemble (ablation) | PASS++ · CIFAR-100 (T=10) | last accuracy (%): 44.69 -> 57.69 | Improved |
| add vanilla rotation ensemble to PODNet (ablation) | PODNet · CIFAR-100 (T=10) | last accuracy (%): 49.45 -> 37.09 | Hurt |
| add vanilla rotation ensemble to DER (ablation) | DER · CIFAR-100 (T=10) | last accuracy (%): 57.38 -> 37.86 | Hurt |
| use soft-label mixing instead of hard label in hardness-aware protoAug (ablation) | PASS++ · CIFAR-100 (T=10) | last accuracy (%) | Hurt |
Closest papers by meaning
- Gradient Reweighting: Towards Imbalanced Class-Incremental Learning (CVPR 2024 · similarity 0.89)
- Long-Tail Class Incremental Learning via Independent Sub-prototype Construction (CVPR 2024 · similarity 0.88)
- Reducing Class-wise Confusion for Incremental Learning with Disentangled Manifolds (CVPR 2025 · similarity 0.87)
- Non-exemplar Online Class-Incremental Continual Learning via Dual-Prototype Self-Augment and Refinement (AAAI 2024 · similarity 0.87)
- eTag: Class-Incremental Learning via Embedding Distillation and Task-Oriented Generation (AAAI 2024 · similarity 0.87)
- Prospective Representation Learning for Non-Exemplar Class-Incremental Learning (NeurIPS 2024 · similarity 0.87)
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.
