Invariant Meta Learning for Out-of-Distribution Generalization
Abstract
Modern deep learning techniques have illustrated their excellent capabilities in many areas, but relies on large training data. Optimization-based meta-learning train a model on a variety tasks, such that it can solve new learning tasks using only a small number of training samples.However, these methods assumes that training and test dataare identically and independently distributed. To overcome such limitation, in this paper, we propose invariant meta learning for out-of-distribution tasks. Specifically, invariant meta learning find invariant optimal meta-initialization,and fast adapt to out-of-distribution tasks with regularization penalty. Extensive experiments demonstrate the effectiveness of our proposed invariant meta learning on out-of-distribution few-shot tasks.
Cite
@article{arxiv.2301.11779,
title = {Invariant Meta Learning for Out-of-Distribution Generalization},
author = {Penghao Jiang and Ke Xin and Zifeng Wang and Chunxi Li},
journal= {arXiv preprint arXiv:2301.11779},
year = {2023}
}
Comments
IEEE Conference on Computer Vision and Pattern Recognition 2022 The Ninth Workshop on Fine-Grained Visual Categorization