Research map / Learning theory and algorithms / Prediction and decision losses
How Important is the Train-Validation Split in 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 |
|---|---|---|---|
| Use all per-task data for both adaptation and evaluation rather than splitting it | Centroid meta-learning with a ridge solver · miniImageNet 1-shot 5-way | accuracy: 48.76 ± 0.87 -> 50.77 ± 0.90 | Helped |
| Use all per-task data for both adaptation and evaluation rather than splitting it | Centroid meta-learning with a ridge solver · miniImageNet 5-shot 5-way | accuracy: 63.56 ± 0.95 -> 67.43 ± 0.89 | Helped |
| Use all per-task data for both adaptation and evaluation rather than splitting it | Centroid meta-learning with a ridge solver · miniImageNet 1-shot 20-way | accuracy: 17.52 ± 0.49 -> 21.17 ± 0.38 | Helped |
| Use all per-task data for both adaptation and evaluation rather than splitting it | Centroid meta-learning with a ridge solver · miniImageNet 5-shot 20-way | accuracy: 21.32 ± 0.54 -> 34.30 ± 0.41 | Helped |
| Use all per-task data for both adaptation and evaluation rather than splitting it | Centroid meta-learning with a ridge solver · tieredImageNet 1-shot 5-way | accuracy: 50.61 ± 1.12 -> 54.37 ± 0.93 | Helped |
| Use all per-task data for both adaptation and evaluation rather than splitting it | Centroid meta-learning with a ridge solver · tieredImageNet 5-shot 5-way | accuracy: 67.30 ± 0.98 -> 71.45 ± 0.94 | Helped |
| Use all per-task data for both adaptation and evaluation rather than splitting it | Centroid meta-learning with a ridge solver · tieredImageNet 1-shot 10-way | accuracy: 29.18 ± 0.57 -> 35.56 ± 0.60 | Helped |
| Use all per-task data for both adaptation and evaluation rather than splitting it | Centroid meta-learning with a ridge solver · tieredImageNet 5-shot 10-way | accuracy: 43.15 ± 0.72 -> 54.50 ± 0.71 | Helped |
| Use train-train instead of train-val on realizable linear tasks with optimally tuned regularization and split ratio | Linear centroid meta-learning · Realizable linear model, high-dimensional limit | MSE constant | Helped |
| Use train-val rather than train-train when task distributions are agnostic | Linear centroid meta-learning · Agnostic task distributions | expected loss alignment with meta-test loss | Helped |
Closest papers by meaning
- Understanding Train-Validation Split in Meta-Learning with Neural Networks (ICLR 2023 · similarity 0.94)
- A Representation Learning Perspective on the Importance of Train-Validation Splitting in Meta-Learning (ICML 2021 · similarity 0.92)
- Crossing the Validation crisis: Cross-validation reduces benchmarking variance surprisingly well (NeurIPS 2026 · similarity 0.85)
- Improving Generalization via Meta-Learning on Hard Samples (CVPR 2024 · similarity 0.85)
- Provable Meta-Learning with Low-Rank Adaptations (NeurIPS 2025 · similarity 0.84)
- MetaOptimize: A Framework for Optimizing Step Sizes and Other Meta-parameters (ICML 2025 · similarity 0.84)
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.