CMVAE: Causal Meta VAE for Unsupervised Meta-Learning

TL;DR

This paper introduces CMVAE, a causal meta-learning model that removes context-bias in unsupervised meta-learning by modeling priors within a causal framework, leading to improved few-shot classification.

Contribution

The paper proposes a novel causal modeling approach for unsupervised meta-learning, explicitly addressing hidden confounders to enhance task adaptation.

Findings

Outperforms state-of-the-art methods on benchmark datasets

Effectively removes context-bias in unsupervised meta-learning

Demonstrates robustness across toy and real datasets

Abstract

Unsupervised meta-learning aims to learn the meta knowledge from unlabeled data and rapidly adapt to novel tasks. However, existing approaches may be misled by the context-bias (e.g. background) from the training data. In this paper, we abstract the unsupervised meta-learning problem into a Structural Causal Model (SCM) and point out that such bias arises due to hidden confounders. To eliminate the confounders, we define the priors are \textit{conditionally} independent, learn the relationships between priors and intervene on them with casual factorization. Furthermore, we propose Causal Meta VAE (CMVAE) that encodes the priors into latent codes in the causal space and learns their relationships simultaneously to achieve the downstream few-shot image classification task. Results on toy datasets and three benchmark datasets demonstrate that our method can remove the context-bias and it outperforms other state-of-the-art unsupervised meta-learning algorithms because of bias-removal. Code is available at \url{https://github.com/GuodongQi/CMVAE}