Research map / Continual learning / Online and replay-based
Adaptive Compositional Continual Meta-Learning
Read the paperCheck 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 |
|---|---|---|---|
| break one-to-one task-component mapping via IBP compositional meta-knowledge (ACML vs mutually exclusive mixture baselines) | continual meta-learning · sequential heterogeneous tasks (Table 1, final Omniglot stage) | mean meta-test accuracy (%): ACML 71.46 vs VC-BML 69.48, DPMM 68.26 | Helped |
| compositional meta-knowledge with evidential sparsification (ACML) | continual meta-learning · mixed-task sequential setting (Table 2, miniImagenet) | average meta-test accuracy (%): 40.04 vs 38.65 (VC-BML) | Helped |
| evidential sparsification removing redundant components (ablation) | ACML · CIFAR-FS stage | meta-test accuracy (%): 86.53 vs 85.99 | Helped |
| evidential sparsification removing redundant components (ablation) | ACML · Omniglot stage | meta-test accuracy (%): 99.43 vs 99.31 | Helped |
| evidential sparsification removing redundant components (ablation) | ACML · miniImagenet stage | meta-test accuracy (%): 75.91 vs 76.07 | No clear effect |
| evidential sparsification removing redundant components (ablation) | ACML · VGG-Flowers stage | meta-test accuracy (%): 71.22 vs 71.41 | No clear effect |
| increasing number of mixture components from 3 to 4 (ablation) | ACML · continual meta-learning across four stages (Fig. 4) | meta-test accuracy (%) | No clear effect |
| increasing number of mixture components from 1 to 3 (ablation) | ACML · continual meta-learning across four stages (Fig. 4) | meta-test accuracy (%) | Helped |
Closest papers by meaning
- Variational Continual Bayesian Meta-Learning (NeurIPS 2021 · similarity 0.91)
- On the Stability-Plasticity Dilemma in Continual Meta-Learning: Theory and Algorithm (NeurIPS 2023 · similarity 0.89)
- MoEMeta: Mixture-of-Experts Meta Learning for Few-Shot Relational Learning (NeurIPS 2025 · similarity 0.87)
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · similarity 0.86)
- Learning to Learn with Contrastive Meta-Objective (NeurIPS 2025 · similarity 0.86)
- Learn more, but bother less: parameter efficient continual learning (NeurIPS 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.