Shallow Bayesian Meta Learning for Real-World Few-Shot Recognition

TL;DR

This paper introduces MetaQDA, a Bayesian meta-learning approach for the classifier layer in few-shot recognition, offering fast, feature-agnostic, and uncertainty-aware classification suitable for real-world applications.

Contribution

It proposes MetaQDA, a novel Bayesian meta-learning method for classifiers that is independent of feature representations and enhances robustness and uncertainty calibration.

Findings

MetaQDA achieves robust cross-domain few-shot performance.

It improves uncertainty calibration in predictions.

The method is fast and memory-efficient.

Abstract

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.