Research map / Label-efficient and robust learning / Zero-/few-shot and meta-learning
Transferring Knowledge across Learning Processes
Sebastian Flennerhag, Pablo G. Moreno, Neil D. Lawrence, Andreas Damianou
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 |
|---|---|---|---|
| Use Leap instead of Reptile | meta-learning across Omniglot alphabets · Omniglot held-out tasks, 25 pretraining tasks | test error: 29.4 -> 24.8 | Improved |
| Use a loss-surface task manifold instead of parameter space alone (ablation) | Leap · Omniglot | prediction error | Improved |
| Use gradient-path length rather than gradient-path energy (ablation) | Leap · Omniglot | performance | Improved |
| Add the stabilizer to the meta objective (ablation) | Leap · Omniglot | convergence speed | Improved |
| Remove the stabilizer (ablation) | Leap · Omniglot | convergence speed | Hurt |
| Use parameter space alone and gradient-path energy, as in the Reptile-like configuration (ablation) | Leap · Omniglot | performance | Hurt |
| Use only one epoch of task training during meta-training (ablation) | Leap · Multi-CV | performance | No clear effect |
Closest papers by meaning
- Theory-Based Causal Transfer:Integrating Instance-Level Induction and Abstract-Level Structure Learning (AAAI 2020 · similarity 0.88)
- Transfer Learning in Deep Reinforcement Learning: A Survey (TPAMI 2023 · similarity 0.88)
- When is Transfer Learning Possible? (ICML 2024 · similarity 0.87)
- Principled Fast and Meta Knowledge Learners for Continual Reinforcement Learning (ICLR 2026 · similarity 0.87)
- Transfer Reinforcement Learning with Shared Dynamics (AAAI 2017 · similarity 0.86)
- Skill-based Meta-Reinforcement Learning (ICLR 2022 · 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.
