Interventional Few-Shot Learning
Zhongqi Yue, Hanwang Zhang, Qianru Sun, Xian-Sheng Hua
Abstract
We uncover an ever-overlooked deficiency in the prevailing Few-Shot Learning (FSL) methods: the pre-trained knowledge is indeed a confounder that limits the performance. This finding is rooted from our causal assumption: a Structural Causal Model (SCM) for the causalities among the pre-trained knowledge, sample features, and labels. Thanks to it, we propose a novel FSL paradigm: Interventional Few-Shot Learning (IFSL). Specifically, we develop three effective IFSL algorithmic implementations based on the backdoor adjustment, which is essentially a causal intervention towards the SCM of many-shot learning: the upper-bound of FSL in a causal view. It is worth noting that the contribution of IFSL is orthogonal to existing fine-tuning and meta-learning based FSL methods, hence IFSL can improve all of them, achieving a new 1-/5-shot state-of-the-art on miniImageNet, tieredImageNet, and cross-domain CUB. Code is released at https://github.com/yue-zhongqi/ifsl.
BibTeX
@inproceedings{NEURIPS2020_1cc8a8ea,
author = {Yue, Zhongqi and Zhang, Hanwang and Sun, Qianru and Hua, Xian-Sheng},
booktitle = {Advances in Neural Information Processing Systems},
editor = {H. Larochelle and M. Ranzato and R. Hadsell and M.F. Balcan and H. Lin},
pages = {2734--2746},
publisher = {Curran Associates, Inc.},
title = {Interventional Few-Shot Learning},
url = {https://proceedings.neurips.cc/paper_files/paper/2020/file/1cc8a8ea51cd0adddf5dab504a285915-Paper.pdf},
volume = {33},
year = {2020}
}