Research map / Continual learning / Online and replay-based
Online Continual Learning on a Contaminated Data Stream With Blurry Task Boundaries
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 PuriDivER instead of the strongest reported baseline | online blurry continual learning · CIFAR-10 SYM-40% | last test accuracy: 50.9 -> 59.2 | Helped |
| Use PuriDivER instead of the strongest reported baseline | online blurry continual learning · CIFAR-100 SYM-60% | last test accuracy: 14.8 -> 28.8 | Helped |
| Use PuriDivER instead of the strongest reported baseline | online blurry continual learning · WebVision | last validation accuracy: 23.1 -> 25.8 | Helped |
| Use PuriDivER instead of the strongest reported baseline | online blurry continual learning · Food-101N | last validation accuracy: 12.0 -> 13.4 | Helped |
| Add re-labeling and consistency regularization together (ablation) | PuriDivER memory training · CIFAR-10 SYM-60% | last test accuracy: 46.9 -> 52.4 | Helped |
| Add consistency regularization to re-labeling (ablation) | PuriDivER memory training · CIFAR-10 SYM-60% | last test accuracy: 44.1 -> 52.4 | Helped |
| Add re-labeling to consistency regularization (ablation) | PuriDivER memory training · CIFAR-10 SYM-60% | last test accuracy: 31.3 -> 52.4 | Helped |
| Use re-labeling alone instead of neither robust-learning component (ablation) | PuriDivER memory training · CIFAR-10 SYM-20% | last test accuracy: 61.8 -> 57.7 | Hurt |
| Use consistency regularization alone instead of neither robust-learning component (ablation) | PuriDivER memory training · CIFAR-10 SYM-20% | last test accuracy: 61.8 -> 48.6 | Hurt |
| Replace adaptive α with static α=0.6 (ablation) | PuriDivER memory sampling · CIFAR-10 SYM-20% | last test accuracy: 29.5 | Hurt |
| Replace adaptive α with static α=1.0 (ablation) | PuriDivER memory sampling · CIFAR-10 SYM-20% | last test accuracy: 26.1 | Hurt |
| Use SELFIE with memory sampling | RSV, GBS, or RM · online noisy-label continual learning | last accuracy | No clear effect |
Closest papers by meaning
- Continual Learning on a Diet: Learning from Sparsely Labeled Streams Under Constrained Computation (ICLR 2024 · similarity 0.89)
- Summarizing Stream Data for Memory-Constrained Online Continual Learning (AAAI 2024 · similarity 0.88)
- Budgeted Online Continual Learning by Adaptive Layer Freezing and Frequency-based Sampling (ICLR 2025 · similarity 0.87)
- Cut out and Replay: A Simple yet Versatile Strategy for Multi-Label Online Continual Learning (ICML 2025 · similarity 0.87)
- Dynamic Sub-graph Distillation for Robust Semi-supervised Continual Learning (AAAI 2024 · similarity 0.87)
- Online Task-Free Continual Learning via Dynamic Expansionable Memory Distribution (CVPR 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.