Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Variational Metric Scaling for Metric-Based 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 |
|---|---|---|---|
| Add D-SVS to PN with cosine similarity | PN, ResNet-12 · miniImageNet test, 5-way 1-shot | accuracy: 52.31±0.83 -> 56.09± 0.19 | Helped |
| Add D-SVS to TADAM | TADAM, ResNet-12 · miniImageNet test, 5-way 1-shot | accuracy: 58.50±0.30 -> 60.16± 0.47 | Helped |
| Add single-parameter SVS to PN | PN, Conv-4 · miniImageNet, 5-way 1-shot, 5-way training | accuracy: 44.15± 0.39 -> 47.84± 0.16 | Helped |
| Replace the single scaling parameter with a dimensional scaling vector | PN + SVS, Conv-4 · miniImageNet, 5-way 1-shot | accuracy: 47.84± 0.16 -> 49.01± 0.39 | Helped |
| Generate task-dependent dimensional scaling vectors | PN + D-SVS, Conv-4 · miniImageNet, 5-way 1-shot | accuracy: 49.01± 0.39 -> 49.10±0.14 | Helped |
| Learn σ rather than fixing it at 0.2 (ablation) | PN + SVS, Conv-4 · miniImageNet, 5-way 1-shot, 5-way training | accuracy: 47.84± 0.16 -> 48.28±0.14 | No clear effect |
| Remove auxiliary training (ablation) | PN + D-AVS, Conv-4 · miniImageNet, 5-way 1-shot | accuracy: 49.10±0.14 -> 48.12± 0.55 | Hurt |
| Remove prior information (ablation) | PN + D-AVS, Conv-4 · miniImageNet, 5-way 1-shot | accuracy: 49.10±0.14 -> 48.56±0.44 | Hurt |
| Remove both auxiliary training and prior information (ablation) | PN + D-AVS, Conv-4 · miniImageNet, 5-way 1-shot | accuracy: 49.10±0.14 -> 47.79± 0.10 | Hurt |
| Use variational SVS instead of training the scaling parameter together with model parameters | PN, Conv-4 · miniImageNet, 5-way 5-shot | accuracy: 66.45 vs. 63.59 | Helped |
Closest papers by meaning
- Learning Low-Dimensional Metrics (NeurIPS 2017 · similarity 0.88)
- Meta-Amortized Variational Inference and Learning (AAAI 2020 · similarity 0.86)
- TADAM: Task dependent adaptive metric for improved few-shot learning (NeurIPS 2018 · similarity 0.86)
- Meta-Learning with Latent Embedding Optimization (ICLR 2019 · similarity 0.86)
- MetaModulation: Learning Variational Feature Hierarchies for Few-Shot Learning with Fewer Tasks (ICML 2023 · similarity 0.85)
- Metric Space Magnitude for Evaluating the Diversity of Latent Representations (NeurIPS 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.