Research map / Optimization / Optimizers
SHOT: Suppressing the Hessian along the Optimization Trajectory for Gradient-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 SHOT_r regularization | MAML, 4-Conv · miniImageNet (5-shot) | accuracy: 64.81 ± 1.63 -> 66.86 ± 0.58 | Helped |
| Add SHOT_r regularization | MAML, 4-Conv · tieredImageNet (5-shot) | accuracy: 66.12 ± 1.10 -> 69.08 ± 0.31 | Helped |
| Add SHOT_r regularization | MAML, 4-Conv · Cars (5-shot) | accuracy: 62.24 ± 2.01 -> 64.20 ± 1.58 | Helped |
| Add SHOT_r regularization | MAML, 4-Conv · CUB (5-shot) | accuracy: 72.48 ± 0.86 -> 73.38 ± 0.32 | Helped |
| Add SHOT_r regularization | MAML, ResNet-12 · Cars (1-shot) | accuracy: 32.74 ± 0.48 -> 32.66 ± 0.27 | Hurt |
| Add SHOT_r regularization | BOIL, 4-Conv · miniImageNet (5-shot) | accuracy: 66.45 ± 0.37 -> 66.34 ± 0.21 | Hurt |
| Replace KL distance with cross-entropy (ablation) | SHOT on MAML, 4-Conv · miniImageNet (5-shot) | accuracy: 64.81 ± 1.63 -> 66.34 ± 0.35 | Helped |
| Replace KL distance with L2 distance (ablation) | SHOT on MAML, 4-Conv · miniImageNet (5-shot) | accuracy: 64.81 ± 1.63 -> 66.85 ± 0.44 | Helped |
| Use L2 distance instead of KL divergence (ablation) | SHOT on MAML, 4-Conv · miniImageNet (5-shot) | accuracy: 66.86 ± 0.58 -> 66.85 ± 0.44 | No clear effect |
| Set λ to 10^-2 (ablation) | SHOT on MAML, 4-Conv · miniImageNet (5-shot) | accuracy: 64.81 ± 1.63 -> 65.97 ± 0.93 | Helped |
| Set λ to 10^-3 (ablation) | SHOT on MAML, 4-Conv · miniImageNet (5-shot) | accuracy: 64.81 ± 1.63 -> 66.85 ± 0.47 | Helped |
| Increase reference optimization steps from 3 to 6 (ablation) | SHOT_3 on MAML, 4-Conv · miniImageNet (5-shot) | accuracy: 66.86 ± 0.58 -> 66.27 ± 0.23 | Hurt |
Closest papers by meaning
- Optimization as a Model for Few-Shot Learning (ICLR 2017 · similarity 0.91)
- MetaDiff: Meta-Learning with Conditional Diffusion for Few-Shot Learning (AAAI 2024 · similarity 0.89)
- A Nested Bi-level Optimization Framework for Robust Few Shot Learning (AAAI 2022 · similarity 0.89)
- MetaNODE: Prototype Optimization as a Neural ODE for Few-Shot Learning (AAAI 2022 · similarity 0.89)
- On Enforcing Better Conditioned Meta-Learning for Rapid Few-Shot Adaptation (NeurIPS 2022 · similarity 0.89)
- Global Convergence of MAML and Theory-Inspired Neural Architecture Search for Few-Shot Learning (CVPR 2022 · similarity 0.88)
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.