Shallow Bayesian Meta Learning for Real-World Few-Shot Recognition
This work provides a fast and memory-efficient meta-learning approach for practitioners in few-shot recognition, particularly for real-world applications requiring better uncertainty calibration.
This paper introduces MetaQDA, a Bayesian meta-learning approach that generalizes quadratic discriminant analysis to meta-learn the classifier layer for few-shot recognition. It achieves robust performance in cross-domain few-shot learning and improves uncertainty calibration in predictions.
Current state-of-the-art few-shot learners focus on developing effective training procedures for feature representations, before using simple, e.g. nearest centroid, classifiers. In this paper, we take an orthogonal approach that is agnostic to the features used and focus exclusively on meta-learning the actual classifier layer. Specifically, we introduce MetaQDA, a Bayesian meta-learning generalization of the classic quadratic discriminant analysis. This setup has several benefits of interest to practitioners: meta-learning is fast and memory-efficient, without the need to fine-tune features. It is agnostic to the off-the-shelf features chosen and thus will continue to benefit from advances in feature representations. Empirically, it leads to robust performance in cross-domain few-shot learning and, crucially for real-world applications, it leads to better uncertainty calibration in predictions.