Effective Meta-Regularization by Kernelized Proximal Regularization

Abstract

We study the problem of meta-learning, which has proved to be advantageous to accelerate learning new tasks with a few samples. The recent approaches based on deep kernels achieve the state-of-the-art performance. However, the regularizers in their base learners are not learnable. In this paper, we propose an algorithm called MetaProx to learn a proximal regularizer for the base learner. We theoretically establish the convergence of MetaProx. Experimental results confirm the advantage of the proposed algorithm.

Cite

Text

Jiang et al. "Effective Meta-Regularization by Kernelized Proximal Regularization." Neural Information Processing Systems, 2021.

Markdown

[Jiang et al. "Effective Meta-Regularization by Kernelized Proximal Regularization." Neural Information Processing Systems, 2021.](https://mlanthology.org/neurips/2021/jiang2021neurips-effective/)

BibTeX

@inproceedings{jiang2021neurips-effective,
  title     = {{Effective Meta-Regularization by Kernelized Proximal Regularization}},
  author    = {Jiang, Weisen and Kwok, James T. and Zhang, Yu},
  booktitle = {Neural Information Processing Systems},
  year      = {2021},
  url       = {https://mlanthology.org/neurips/2021/jiang2021neurips-effective/}
}