Research map / Label-efficient and robust learning / Domain adaptation and generalization
Learning To Learn and Remember Super Long Multi-Domain Task Sequence
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 the meta optimizer and adaptive freezing with PNet | PNet · SDML 10-domain benchmark, 5-way 5-shot, domain-aware | ACC: 50.79 ± 0.37 -> 55.28 ± 0.19 | Helped |
| Use the meta optimizer and adaptive freezing with PNet | PNet · SDML 10-domain benchmark, 5-way 5-shot, domain-aware | BWT: −13.82 ± 0.45 -> −11.15 ± 0.27 | Helped |
| Use the meta optimizer and adaptive freezing with ANIL | ANIL · SDML 10-domain benchmark, 5-way 5-shot, domain-aware | ACC: 48.08 ± 0.33 -> 51.56 ± 0.21 | Helped |
| Use the meta optimizer and adaptive freezing with ANIL | ANIL · SDML 10-domain benchmark, 5-way 5-shot, domain-aware | BWT: −20.30 ± 0.35 -> −16.07 ± 0.20 | Helped |
| Add the meta optimizer and latent domain-shift detection to PNet | PNet · SDML 10-domain benchmark, 5-way 5-shot, domain-agnostic | ACC: 49.64 ± 0.27 -> 54.67 ± 0.20 | Helped |
| Add the meta optimizer and latent domain-shift detection to PNet | PNet · SDML 10-domain benchmark, 5-way 5-shot, domain-agnostic | BWT: −18.29 ± 0.31 -> −11.67 ± 0.28 | Helped |
| Add the meta optimizer and latent domain-shift detection to ANIL | ANIL · SDML 10-domain benchmark, 5-way 5-shot, domain-agnostic | ACC: 48.08 ± 0.33 -> 51.18 ± 0.31 | Helped |
| Add the meta optimizer and latent domain-shift detection to ANIL | ANIL · SDML 10-domain benchmark, 5-way 5-shot, domain-agnostic | BWT: −20.30 ± 0.35 -> −16.85 ± 0.29 | Helped |
| Use the proposed method instead of Continual-MAML | SDML continual meta-learning · SDML 10-domain benchmark, 5-way 1-shot | ACC: 36.36 ± 1.12 -> 40.23 ± 0.32 | Helped |
| Use the proposed method instead of Continual-MAML | SDML continual meta-learning · SDML 10-domain benchmark, 5-way 5-shot | ACC: 49.81 ± 0.89 -> 55.28 ± 0.19 | Helped |
| Increase the number of training domains in the sequence | Proposed SDML method · SDML 5-way 5-shot; domain counts shown in Figure 3 | ACC | Mixed |
| Use the proposed latent space instead of raw task embeddings for domain-shift detection | BOCPD domain-shift detector · SDML domain-shift detection | MIC and TIC dependence analysis | Helped |
Closest papers by meaning
- Addressing Catastrophic Forgetting in Few-Shot Problems (ICML 2021 · similarity 0.89)
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · similarity 0.88)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · similarity 0.87)
- Mitigating Catastrophic Forgetting in Online Continual Learning by Modeling Previous Task Interrelations via Pareto Optimization (ICML 2024 · similarity 0.87)
- Make Domain Shift a Catastrophic Forgetting Alleviator in Class-Incremental Learning (AAAI 2025 · similarity 0.87)
- Unlocking the Power of Function Vectors for Characterizing and Mitigating Catastrophic Forgetting in Continual Instruction Tuning (ICLR 2025 · 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.