Research map / Continual learning / Class-incremental
ACIL: Analytic Class-Incremental Learning with Absolute Memorization and Privacy Protection
HUIPING ZHUANG, Zhenyu Weng, Hongxin Wei, RENCHUNZI XIE, Kar-Ann Toh, Zhiping Lin
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 |
|---|---|---|---|
| use ACIL instead of POD+AANets+RMM | CIFAR-100, 5-phase CIL · CIFAR-100, 5-phase | average incremental accuracy (%): 68.36 -> 66.30 | Hurt |
| use ACIL instead of POD+AANets+RMM | CIFAR-100, 25-phase CIL · CIFAR-100, 25-phase | average incremental accuracy (%): 64.12 -> 65.95 | Improved |
| use ACIL instead of POD+AANets+RMM | ImageNet-Subset, 25-phase CIL · ImageNet-Subset, 25-phase | average incremental accuracy (%): 75.01% v.s. 74.59% | Hurt |
| use ACIL instead of LUCIR+Mnemonics | ImageNet-Full, 25-phase CIL · ImageNet-Full, 25-phase | average incremental accuracy (%): 63.08 -> 64.63 | Improved |
| use ACIL instead of PODNet | CIFAR-100, 50-phase CIL · CIFAR-100, 50-phase | average incremental accuracy (%): 57.98% v.s. 66.01% | Improved |
| add feature expansion (ablation) | ACIL on CIFAR-100, 5-phase · CIFAR-100, 5-phase | average incremental accuracy (%): 52.99% -> 66.30% | Improved |
| add regularization at 10^-1 rather than none (ablation) | ACIL on CIFAR-100, 5-phase · CIFAR-100, 5-phase | average incremental accuracy (%): 51.12% -> 66.30% | Improved |
| set regularization to 10^-2 instead of 10^-1 (ablation) | ACIL on CIFAR-100, 5-phase · CIFAR-100, 5-phase | average incremental accuracy (%): 66.30 -> 66.25 | No clear effect |
| set regularization to 10^-3 instead of 10^-1 (ablation) | ACIL on CIFAR-100, 5-phase · CIFAR-100, 5-phase | average incremental accuracy (%): 66.30 -> 66.23 | No clear effect |
Later papers that compared with this method
3 later accepted papers report ACIL as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Online Continual Learning with Dynamic Label Hierarchies (ICML 2026 · better on 8 of 8 shared results)
- Knowledge Memorization and Rumination for Pre-trained Model-based Class-Incremental Learning (CVPR 2025 · better on 3 of 3 shared results)
- DS-AL: A Dual-Stream Analytic Learning for Exemplar-Free Class-Incremental Learning (AAAI 2024 · better on 18 of 18 shared results)
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Long-Tail Class Incremental Learning via Independent Sub-prototype Construction (CVPR 2024 · similarity 0.87)
- eTag: Class-Incremental Learning via Embedding Distillation and Task-Oriented Generation (AAAI 2024 · similarity 0.86)
- AnaCP: Toward Upper-Bound Continual Learning via Analytic Contrastive Projection (NeurIPS 2025 · similarity 0.86)
- Defying Imbalanced Forgetting in Class Incremental Learning (AAAI 2024 · similarity 0.86)
- GACL: Exemplar-Free Generalized Analytic Continual Learning (NeurIPS 2024 · similarity 0.86)
- Federated Class-Incremental Learning: A Hybrid Approach Using Latent Exemplars and Data-Free Techniques to Address Local and Global Forgetting (ICLR 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.
