Research map / Learning theory and algorithms / Kernels and regression theory
Effective Meta-Regularization by Kernelized Proximal Regularization
Weisen Jiang, James Kwok, Yu Zhang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 learnable proximal regularizer f_theta (MetaProx vs MetaOptNet-RR, f_theta=0) | MetaOptNet-RR · Sine 2-shot noisy | MSE: 0.79 -> 0.43 | Improved |
| add learnable f_theta | MetaOptNet-RR · Sine 2-shot noise-free | MSE: 0.18 -> 0.11 | Improved |
| add learnable f_theta | MetaOptNet-RR · Sine 5-shot noisy | MSE: 0.19 -> 0.13 | Improved |
| add learnable f_theta | MetaOptNet-RR · Sine 5-shot noise-free | MSE: 0.01 -> 0.01 | No clear effect |
| add learnable f_theta | MetaOptNet-RR · Sale 1-shot | MSE: 0.088 -> 0.061 | Improved |
| add learnable f_theta | MetaOptNet-RR · QMUL in-range | MSE: 0.021 -> 0.012 | Improved |
| add learnable f_theta | MetaOptNet-RR · QMUL out-of-range | MSE: 0.039 -> 0.020 | Improved |
| add learnable regularizer f_theta to MetaOptNet-SVM(cos), Conv4 | MetaOptNet-SVM(cos) · mini-ImageNet 5-way 1-shot Conv4 | accuracy: 50.1 -> 52.4 | Improved |
| add learnable regularizer f_theta to MetaOptNet-SVM(cos), ResNet-12 | MetaOptNet-SVM(cos) · mini-ImageNet 5-way 1-shot ResNet-12 | accuracy: 62.75 -> 63.82 | Improved |
| add learnable regularizer f_theta to MetaOptNet-SVM(cos), ResNet-12 | MetaOptNet-SVM(cos) · mini-ImageNet 5-way 5-shot ResNet-12 | accuracy: 78.68 -> 79.12 | Improved |
| use kernel-based methods instead of non-kernel meta-regularization (Meta-MinibatchProx) | MetaProx · QMUL in-range | MSE: 0.171 -> 0.012 | Improved |
| replace linear model CommonMean with kernel/neural methods | MetaProx · Sine 2-shot noise-free | MSE: 4.58 -> 0.11 | Improved |
Closest papers by meaning
- Provable Guarantees for Gradient-Based Meta-Learning (ICML 2019 · similarity 0.87)
- Dynamic Kernel Selection for Improved Generalization and Memory Efficiency in Meta-Learning (CVPR 2022 · similarity 0.87)
- Improving Generalization of Meta-Learning With Inverted Regularization at Inner-Level (CVPR 2023 · similarity 0.87)
- MetaReg: Towards Domain Generalization using Meta-Regularization (NeurIPS 2018 · similarity 0.86)
- Proximal Mapping for Deep Regularization (NeurIPS 2020 · similarity 0.86)
- Adaptive Proximal Gradient Methods for Structured Neural Networks (NeurIPS 2021 · 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.
