Research map / Continual learning / Online and replay-based
Anti-Retroactive Interference for Lifelong Learning
Runqi Wang, Yuxiang Bao, Baochang Zhang, Jianzhuang Liu, Wentao Zhu, Guodong Guo
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 |
|---|---|---|---|
| ARI compared with best prior method RPSnet | RPSnet · ImageNet100 | accuracy: 74.10 -> 79.32 | Improved |
| ARI compared with iTAML after 10 tasks | iTAML · CIFAR100 (10 tasks) | average accuracy: 77.79% -> 80.88% | Improved |
| ARI compared with iTAML at early tasks (task 5) | iTAML · CIFAR100 (10 tasks) | average accuracy: 84.31% -> 83.10% | Hurt |
| add background attack with mask B = 1 - A∘A (ablation) | ARI · CIFAR100 | accuracy | Improved |
| use mask B = 1/A instead of 1 - A∘A (ablation) | ARI background attack · CIFAR100 | accuracy | Hurt |
| use mask B = 1 - A instead of 1 - A∘A (ablation) | ARI background attack · CIFAR100 | accuracy | Hurt |
| add distance-weighted task-specific model fusion (vs average fusion) (ablation) | ARI · CIFAR100 | accuracy | Improved |
| replace backgrounds of training images with similar ones (toy experiment) (ablation) | toy 10-class dataset · toy dataset | average accuracy and BWT | Hurt |
| ARI memory-based lifelong learning versus GEM on forgetting | GEM · CIFAR100 | BWT: 1.2% -> -7.5% | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- iTAML — iTAML: An Incremental Task-Agnostic Meta-learning Approach (CVPR 2020) · better on 3 of 5 shared results
- iCARL — iCaRL: Incremental Classifier and Representation Learning (CVPR 2017) · better on 3 of 3 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
- Principled Fast and Meta Knowledge Learners for Continual Reinforcement Learning (ICLR 2026 · similarity 0.87)
- Memory Aware Synapses: Learning what (not) to forget (ECCV 2018 · similarity 0.87)
- Online Fast Adaptation and Knowledge Accumulation (OSAKA): a New Approach to Continual Learning (NeurIPS 2020 · similarity 0.86)
- On the Stability-Plasticity Dilemma in Continual Meta-Learning: Theory and Algorithm (NeurIPS 2023 · similarity 0.86)
- A Unified and General Framework for Continual Learning (ICLR 2024 · similarity 0.86)
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · 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.
