Research map / Continual learning / Online and replay-based
Continual Learning on Noisy Data Streams via Self-Purified Replay
Chris Dongjoo Kim, Jinseo Jeong, Sangwoo Moon, Gunhee Kim
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 the Self-Centered filter to Self-Replay (ablation) | Self-Replay · MNIST symmetric noise 60% | accuracy: 43.0 -> 84.8 | Improved |
| add the Self-Centered filter to Self-Replay (ablation) | Self-Replay · CIFAR-10 symmetric noise 60% | accuracy: 22.4 -> 40.0 | Improved |
| add Self-Replay to the Self-Centered filter (ablation) | Self-Centered filter · MNIST symmetric noise 20% | accuracy: 80.1 -> 85.4 | Improved |
| add the Self-Centered filter to Self-Replay (ablation) | Self-Replay · WebVision (real noise) | accuracy: 48.0 -> 40.0 | Hurt |
| add stochastic ensemble to the Self-Centered filter (ablation) | non-stochastic Self-Centered filter · MNIST symmetric noise 20% (filtered noise %) | filtered noisy label percentage: 79.5 -> 96.0 | Improved |
| add stochastic ensemble to the Self-Centered filter (ablation) | non-stochastic Self-Centered filter · CIFAR-10 symmetric noise 20% (filtered noise %) | filtered noisy label percentage: 50.5 -> 75.5 | Improved |
| use the Self-Centered filter instead of AUM | AUM · CIFAR-10 symmetric noise 20% (filtered noise %) | filtered noisy label percentage: 36.0 -> 75.5 | Improved |
| add L2R noisy label technique | GDumb · MNIST symmetric noise 20% | accuracy: 70.0 -> 65.2 | Hurt |
| add L2R noisy label technique | GDumb · MNIST symmetric noise 60% | accuracy: 36.0 -> 42.3 | Improved |
| add L2R noisy label technique | CRS · MNIST symmetric noise 20% | accuracy: 58.6 -> 80.6 | Improved |
| use SPR instead of GDumb + SL | GDumb + SL · CIFAR-100 random symmetric noise 20% | accuracy: 19.3 -> 21.5 | Improved |
| use SPR instead of GDumb + JoCoR | GDumb + JoCoR · CIFAR-100 random symmetric noise 60% | accuracy: 6.1 -> 18.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- GDumb — GDumb: A Simple Approach that Questions Our Progress in Continual Learning (ECCV 2020) · better on 7 of 7 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
- Predicting the Susceptibility of Examples to Catastrophic Forgetting (ICML 2025 · similarity 0.90)
- Layerwise Proximal Replay: A Proximal Point Method for Online Continual Learning (ICML 2024 · similarity 0.89)
- Summarizing Stream Data for Memory-Constrained Online Continual Learning (AAAI 2024 · similarity 0.89)
- Cut out and Replay: A Simple yet Versatile Strategy for Multi-Label Online Continual Learning (ICML 2025 · similarity 0.89)
- Continual Learning on a Diet: Learning from Sparsely Labeled Streams Under Constrained Computation (ICLR 2024 · similarity 0.89)
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.88)
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.
