Research map / Continual learning / Online and replay-based
Co$^{2}$ In : a Bi-level Memory Incremental Learning Framework with Knowledge Encoding, Consolidation, and Integration
Wenju Sun, Qingyong Li, Boyang Li, Wen Wang, Siyu Zhang, Yi Wei, Yangli‐ao Geng
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 representation compaction regularizer (ablation) | Co2In · CIFAR-100 Class-IL | accuracy: 34.75 -> 37.54 | Improved |
| add basis updating in knowledge consolidation (ablation) | Co2In · CIFAR-100 Class-IL | accuracy: 6.88 -> 37.54 | Improved |
| add basis updating in knowledge consolidation (ablation) | Co2In · Tiny-ImageNet Class-IL | accuracy: 2.90 -> 31.23 | Improved |
| add retraining of knowledge representation after projection (ablation) | Co2In · CIFAR-100 Class-IL | accuracy: 36.12 -> 37.54 | Improved |
| replace linear classifier with quadratic classifier (ablation) | Co2In · CIFAR-100 Class-IL | accuracy: 16.90 -> 37.54 | Improved |
| replace linear classifier with quadratic classifier (ablation) | Co2In · Tiny-ImageNet Class-IL | accuracy: 8.81 -> 31.23 | Improved |
| replace linear classifier with quadratic classifier (ablation) | Co2In · Tiny-ImageNet Task-IL | accuracy: 69.31 -> 68.75 | No clear effect |
| add quadratic classifier to GPM (ablation) | GPM · CIFAR-100 Class-IL | accuracy: 8.19 -> 17.04 | Improved |
| add quadratic classifier to GPM (ablation) | GPM · CIFAR-100 Task-IL | accuracy: 70.93 -> 61.11 | Hurt |
| add quadratic classifier to LwF and DER++ (ablation) | LwF, DER++ · CIFAR-100 | accuracy: FAIL | Hurt |
| add rotate class augmentation with 50-task split (ablation) | Co2In · CIFAR-100 Class-IL (50 tasks) | accuracy: 5.02 -> 21.73 | Improved |
| add rotate class augmentation on 10-task split (ablation) | Co2In · CIFAR-100 Class-IL (10 tasks) | accuracy: 37.54 -> 29.58 | Hurt |
Closest papers by meaning
- Rethinking the Stability-Plasticity Trade-off in Continual Learning from an Architectural Perspective (ICML 2025 · similarity 0.84)
- An Efficient Memory Module for Graph Few-Shot Class-Incremental Learning (NeurIPS 2024 · similarity 0.84)
- Task-Free Continual Generation and Representation Learning via Dynamic Expansionable Memory Cluster (AAAI 2024 · similarity 0.83)
- MIND: Multi-Task Incremental Network Distillation (AAAI 2024 · similarity 0.83)
- Multi-View Incremental Learning with Structured Hebbian Plasticity for Enhanced Fusion Efficiency (AAAI 2025 · similarity 0.83)
- Bisecle: Binding and Separation in Continual Learning for Video Language Understanding (NeurIPS 2025 · similarity 0.83)
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.
