Research map / Probabilistic and scientific modeling / Variational and generative sampling
MetaNODE: Prototype Optimization as a Neural ODE for Few-Shot 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 |
|---|---|---|---|
| Use prototype optimization with MetaNODE instead of the strongest reported baseline | Few-shot classification · miniImagenet 5-way 1-shot, transductive | accuracy: 73.13 ± 0.85% -> 77.92 ± 0.99% | Helped |
| Use prototype optimization with MetaNODE instead of the strongest reported baseline | Few-shot classification · tieredImagenet 5-way 1-shot, transductive | accuracy: 80.56 % -> 83.46 ± 0.92% | Helped |
| Use MetaNODE instead of the strongest reported baseline | Inductive few-shot classification · miniImagenet 5-way 1-shot | accuracy: 64.22±0.67% -> 66.07 ± 0.79% | Helped |
| Use MetaNODE instead of the strongest reported baseline | Inductive few-shot classification · CUB-200-2011 5-way 1-shot | accuracy: 74.96 ± 0.86% -> 80.82 ± 0.75% | Helped |
| Add a prototype meta-optimizer (ablation) | Inductive MetaNODE framework · miniImagenet 5-way 1-shot | accuracy: 61.22 ± 0.84% -> 63.85 ± 0.81%; 61.22 ± 0.84% -> 64.37 ± 0.79% | Helped |
| Replace MetaLSTM with Neural ODE (ablation) | Inductive prototype optimization · miniImagenet 5-way 1-shot | accuracy: 63.85 ± 0.81% -> 66.07 ± 0.79%; 64.37 ± 0.79% -> 66.07 ± 0.79% | Helped |
| Add unlabeled query samples to Neural ODE (ablation) | MetaNODE · miniImagenet 5-way 1-shot | accuracy: 66.07 ± 0.79% -> 77.92 ± 0.99% | Helped |
| Remove ensemble mechanism (ablation) | MetaNODE GradNet · miniImagenet 5-way 1-shot | accuracy: 77.92 ± 0.99% -> 75.34 ± 1.10% | Hurt |
| Remove attention mechanism (ablation) | MetaNODE GradNet · miniImagenet 5-way 1-shot | accuracy: 77.92 ± 0.99% -> 75.10 ± 0.98% | Hurt |
| Remove exponential decay (ablation) | MetaNODE GradNet · miniImagenet 5-way 1-shot | accuracy: 77.92 ± 0.99% -> 76.02 ± 1.17% | Hurt |
| Infer gradient flow with GradNet rather than average support gradients | MetaNODE · miniImagenet gradient bias analysis | cosine similarity to real gradient: 0.0441 -> 0.1701 | Helped |
Closest papers by meaning
- MetaDiff: Meta-Learning with Conditional Diffusion for Few-Shot Learning (AAAI 2024 · similarity 0.89)
- SHOT: Suppressing the Hessian along the Optimization Trajectory for Gradient-Based Meta-Learning (NeurIPS 2023 · similarity 0.89)
- Optimization as a Model for Few-Shot Learning (ICLR 2017 · similarity 0.89)
- Prototype Completion With Primitive Knowledge for Few-Shot Learning (CVPR 2021 · similarity 0.88)
- Meta-Learning with Implicit Gradients (NeurIPS 2019 · similarity 0.88)
- Memory-Reduced Meta-Learning with Guaranteed Convergence (AAAI 2025 · 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.