Research map / Learning theory and algorithms / Kernels and regression theory
Towards Sample-efficient Overparameterized 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 an optimally eigen-weighted representation instead of an unweighted projection | Overparameterized linear meta-learning · Few-shot linear regression | few-shot risk | Helped |
| Increase representation dimension beyond the few-shot sample count while using optimal eigen-weighting | Overparameterized linear meta-learning · Few-shot linear regression | few-shot risk | Helped |
| Remove eigen-weighting, using a simple projection (ablation) | Overparameterized linear meta-learning · Few-shot linear regression | few-shot risk | Hurt |
| Apply eigen-weighting in the underparameterized regime (ablation) | Linear meta-learning · Few-shot linear regression | few-shot risk | No clear effect |
| Estimate the representation from a finite number of earlier-task samples rather than using the true covariance | Overparameterized linear meta-learning · Few-shot linear regression | few-shot risk | Hurt |
| Increase representation dimension R | Estimated-representation meta-learning · Few-shot linear regression | risk | Mixed |
| Use feature covariance as an inductive bias when estimating the canonical task covariance | MoM representation learning · Representation learning | canonical-subspace estimation error | Helped |
| Use n_1 = s_T samples per task instead of n_1 = 1 in the identity-feature, low-rank-task setting (ablation) | MoM · MoM estimation, identity feature covariance and low-rank task covariance | sample complexity N: ds_T^2 -> ds_T | Helped |
| Use n_1 = r_T samples per task instead of n_1 = 1 in the general spiked-covariance setting (ablation) | MoM · MoM estimation, general spiked feature and task covariance | sample complexity N: r_F r_T^2 -> r_Fr_T | Helped |
Closest papers by meaning
- Theoretical Characterization of the Generalization Performance of Overfitted Meta-Learning (ICLR 2023 · similarity 0.92)
- Understanding Benign Overfitting in Gradient-Based Meta Learning (NeurIPS 2022 · similarity 0.91)
- Provable Generalization of Overparameterized Meta-learning Trained with SGD (NeurIPS 2022 · similarity 0.89)
- Meta-Learning Requires Meta-Augmentation (NeurIPS 2020 · similarity 0.88)
- Improving Generalization via Meta-Learning on Hard Samples (CVPR 2024 · similarity 0.87)
- Simplicity Bias in Overparameterized Machine Learning (AAAI 2024 · 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.