Research map / Continual learning / Online and replay-based
Learning from the Past: Continual Meta-Learning with Bayesian Graph Neural Networks
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 |
|---|---|---|---|
| add history transition (gated recurrent memory across episodes) (ablation) | CML-BGNN · miniImageNet 5-way 1-shot | accuracy (%): 63.74 -> 88.62 | Helped |
| add history transition (gated recurrent memory across episodes) (ablation) | CML-BGNN · miniImageNet 5-way 5-shot | accuracy (%): 79.36 -> 92.69 | Helped |
| add Bayes by Backprop task-specific edge inference (ablation) | CML-BGNN · miniImageNet 5-way 5-shot | accuracy (%): 91.21 -> 92.69 | Helped |
| add Bayes by Backprop task-specific edge inference (ablation) | CML-BGNN · miniImageNet 5-way 1-shot | accuracy (%): 87.15 -> 88.62 | Helped |
| increase graph depth from 1 to 3 layers (ablation) | CML-BGNN · miniImageNet 5-way 1-shot | accuracy (%): 85.73 -> 88.62 | Helped |
| increase graph depth from 1 to 3 layers | GNN (re-implemented) · miniImageNet 5-way 5-shot | accuracy (%): 65.58 -> 66.99 | Mixed |
| increase GRU hidden states from 4 or 8 to 16 (ablation) | CML-BGNN · miniImageNet 5-way 1-shot validation | validation accuracy | Helped |
| add unlabeled support data (semi-supervised) at 20% labeled | CML-BGNN · miniImageNet 5-way 5-shot | accuracy (%): 84.37 -> 88.95 | Helped |
| enlarge mini-batch size from 40 to 80 | EGNN · tieredImageNet | accuracy (%) | Helped |
Closest papers by meaning
- Learning to Continually Learn with the Bayesian Principle (ICML 2024 · similarity 0.92)
- Posterior Meta-Replay for Continual Learning (NeurIPS 2021 · similarity 0.90)
- Variational Continual Bayesian Meta-Learning (NeurIPS 2021 · similarity 0.89)
- Addressing Catastrophic Forgetting in Few-Shot Problems (ICML 2021 · similarity 0.89)
- Exploring Rationale Learning for Continual Graph Learning (AAAI 2025 · similarity 0.87)
- Re-evaluating Continual Learning with Few-Shot Adaptation (NeurIPS 2026 · 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.