Research map / Probabilistic and scientific modeling / Variational and generative sampling
Learning to Learn Variational Semantic Memory
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 |
|---|---|---|---|
| model prototypes as distributions via variational inference instead of averaged vectors | ProtoNet · miniImageNet 5-way 1-shot | accuracy (%): 47.40 -> 52.11 | Helped |
| add variational semantic memory (ablation) | Variational prototype network · miniImageNet 5-way 1-shot | accuracy (%): 52.11 -> 54.73 | Helped |
| add variational semantic memory (ablation) | Variational prototype network · CIFAR-FS 5-way 1-shot | accuracy (%): 61.35 -> 63.42 | Helped |
| use variational latent memory recall instead of rote memory (ablation) | VSM · miniImageNet 5-way 1-shot | accuracy (%): 53.15 -> 54.73 | Helped |
| use variational inference recall instead of deterministic transformed memory (ablation) | VSM · miniImageNet 5-way 1-shot | accuracy (%): 53.85 -> 54.73 | Helped |
| use attention-based memory update instead of mean-based (ablation) | VSM · miniImageNet 5-way 1-shot | accuracy (%): 53.97 -> 54.73 | Helped |
| use attention-based memory update instead of mean-based (ablation) | VSM · CIFAR-FS 5-way 5-shot | accuracy (%): 76.79 -> 77.93 | Helped |
| replace softmax addressing with Gumbel-softmax (ablation) | VSM · miniImageNet 5-way 1-shot | accuracy (%): 54.73 -> 53.27 | Mixed |
| vary memory update retention alpha (best at 0.7 to 0.8) (ablation) | VSM · miniImageNet 5-way 1-shot | accuracy (%): 51.28 -> 54.73 | Helped |
| increase memory size (ablation) | VSM · miniImageNet 5-way 1-shot | accuracy (%) | Helped |
| use VSM with ResNet-12 backbone vs prior best | Tian et al. · miniImageNet 5-way 1-shot | accuracy (%): 64.82 -> 65.72 | Helped |
| use VSM vs MANN memory | MANN · miniImageNet 5-way 1-shot | accuracy (%): 41.38 -> 54.73 | Helped |
Closest papers by meaning
- Hierarchical Variational Memory for Few-shot Learning Across Domains (ICLR 2022 · similarity 0.93)
- Variational Memory Addressing in Generative Models (NeurIPS 2017 · similarity 0.87)
- Knowledge Graph Enhanced Generative Multi-modal Models for Class-Incremental Learning (NeurIPS 2025 · similarity 0.85)
- Task-Free Continual Generation and Representation Learning via Dynamic Expansionable Memory Cluster (AAAI 2024 · similarity 0.85)
- Unlocking Compositional Generalization in Continual Few-Shot Learning (NeurIPS 2026 · similarity 0.85)
- Simple Semantic-Aided Few-Shot Learning (CVPR 2024 · similarity 0.85)
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.