Research map / Continual learning / Online and replay-based
Beyond Myopic Alignment: Lookahead Optimization for Online Class-Incremental Learning
Song Lai, Zhe Zhao, Fei Zhu, Ji Cheng, Xi Lin, Qingfu Zhang, Gaofeng Meng
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 |
|---|---|---|---|
| LOR lookahead optimization vs best baseline POCL | LOR · Seq-CIFAR10 (|M|=1k) | AAA (%): 66.23 -> 74.22 | Improved |
| LOR vs POCL | LOR · Seq-CIFAR100 (|M|=1k) | Acc (%): 16.54 -> 18.21 | Mixed |
| LOR vs POCL | LOR · Seq-TinyImageNet (|M|=5k) | Acc (%): 19.40 -> 20.25 | Improved |
| use LOR vs POCL on forgetting | LOR · Seq-CIFAR100 (|M|=5k) | FM: -17.55 -> -11.12 | Improved |
| use LOR on smaller 3-layer DNN | LOR vs ER · Seq-CIFAR10 (3-layer DNN) | AAA: 48.50 -> 64.41 | Improved |
| use LOR on full ResNet-18 | LOR vs VR-MCL · Seq-CIFAR100 (full ResNet-18) | Acc (%): 19.68 -> 25.45 | Improved |
| lengthen task sequence to 40 tasks | LOR vs POCL · Seq-TinyImageNet (40 tasks) | AAA (%): 19.96 -> 24.15 | Improved |
| vary LSE smoothing mu from 0.5 to 0.2 or 2.0 (ablation) | LOR · Seq-CIFAR100 (|M|=5k) | AAA (%): 43.26 -> 42.88 (mu=0.2), 42.31 (mu=2.0) | No clear effect |
| vary Dirichlet concentration alpha_s from 1.0 to 0.5 or 2.0 (ablation) | LOR · Seq-CIFAR100 (|M|=5k) | AAA (%): 43.26 -> 42.95 (0.5), 42.88 (2.0) | No clear effect |
Closest papers by meaning
- Defying Imbalanced Forgetting in Class Incremental Learning (AAAI 2024 · similarity 0.88)
- Multi-layer Rehearsal Feature Augmentation for Class-Incremental Learning (ICML 2024 · similarity 0.88)
- Adapter Merging with Centroid Prototype Mapping for Scalable Class-Incremental Learning (CVPR 2025 · similarity 0.88)
- Long-Tail Class Incremental Learning via Independent Sub-prototype Construction (CVPR 2024 · similarity 0.87)
- Forgetting, Ignorance or Myopia: Revisiting Key Challenges in Online Continual Learning (NeurIPS 2024 · similarity 0.87)
- MOS: Model Surgery for Pre-Trained Model-Based Class-Incremental Learning (AAAI 2025 · 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.
