Research map / Adversarial robustness and security / Certified and empirical robustness
Adversarial Robust Memory-Based Continual Learner
Xiaoyue Mi, Fan Tang, Zonghan Yang, Danding Wang, Juan Cao, Peng Li, Yang Liu
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 anti-forgettable logit calibration (AFLC) (ablation) | ER+TRADES · Split-CIFAR10 (buffer 200, class-incremental) | clean forgetting: 77.25 -> 21.76 | Improved |
| add AFLC (ablation) | ER+TRADES · Split-CIFAR10 (buffer 200, class-incremental) | clean FAA: 22.42 -> 37.88 | Improved |
| add AFLC (ablation) | ER+TRADES · Split-CIFAR10 (buffer 200, task-incremental) | clean forgetting: 8.10 -> 10.86 | Hurt |
| add robustness-aware experience replay (RAER) on top of AFLC (ablation) | ER+TRADES+AFLC · Split-CIFAR10 (buffer 200, class-incremental) | adversarial FAA: 18.58 -> 19.81 | Improved |
| add RAER on top of AFLC (ablation) | ER+TRADES+AFLC · Split-CIFAR10 (buffer 200, task-incremental) | adversarial FAA: 53.64 -> 55.71 | Improved |
| add RAER on top of AFLC (ablation) | ER+TRADES+AFLC · Split-CIFAR10 (buffer 200, class-incremental) | adversarial forgetting: 21.77 -> 9.61 | Improved |
| add future prior adjustment (FP) (ablation) | ER+TRADES+AFLC+RAER · Split-CIFAR10 (buffer 200, class-incremental) | clean FAA: 37.45 -> 43.34 | Improved |
| add future prior adjustment (FP) (ablation) | ER+TRADES+AFLC+RAER · Split-CIFAR10 (buffer 200, class-incremental) | clean forgetting: 10.29 -> 33.40 | Hurt |
| apply RAER data selection threshold rho=5 only (without AFLC) (ablation) | ER+TRADES · Split-CIFAR10 (buffer 200, class-incremental) | clean FAA: 22.42 -> 18.21 | Hurt |
| apply RAER data selection threshold rho=10 only (without AFLC) (ablation) | ER+TRADES · Split-CIFAR10 (buffer 200, task-incremental) | adversarial FAA: 51.33 -> 59.88 | Improved |
| increase calibration alpha from 3.5 to 7.0 (ablation) | AFLC · Split-CIFAR10 (buffer 200, class-incremental) | adversarial FAA: 18.58 -> 13.89 | Hurt |
| add our method (AFLC + RAER) | ER+TRADES · Split-CIFAR10 (buffer 200, task-incremental) | adversarial FAA (PGD-20): 51.33 -> 59.41 | Improved |
Closest papers by meaning
- The Effectiveness of Random Forgetting for Robust Generalization (ICLR 2024 · similarity 0.88)
- Resurrecting Old Classes with New Data for Exemplar-Free Continual Learning (CVPR 2024 · similarity 0.88)
- BrainWash: A Poisoning Attack to Forget in Continual Learning (CVPR 2024 · similarity 0.87)
- Recall-Oriented Continual Learning with Generative Adversarial Meta-Model (AAAI 2024 · similarity 0.87)
- Towards Understanding Catastrophic Forgetting in Two-layer Convolutional Neural Networks (ICML 2025 · similarity 0.87)
- An Effective Dynamic Gradient Calibration Method for Continual Learning (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.
