Few-shot Learning with Noisy Labels

TL;DR

This paper addresses the challenge of label noise in few-shot learning by proposing feature aggregation enhancements and a Transformer-based model, TraNFS, which effectively identifies and mitigates mislabeled samples, improving robustness.

Contribution

It introduces a new Transformer model, TraNFS, and improved feature aggregation methods to enhance robustness of FSL methods against noisy labels.

Findings

TraNFS performs comparably to leading methods on clean data.

TraNFS significantly outperforms others with noisy support sets.

Proposed methods improve robustness to label noise in FSL.

Abstract

Few-shot learning (FSL) methods typically assume clean support sets with accurately labeled samples when training on novel classes. This assumption can often be unrealistic: support sets, no matter how small, can still include mislabeled samples. Robustness to label noise is therefore essential for FSL methods to be practical, but this problem surprisingly remains largely unexplored. To address mislabeled samples in FSL settings, we make several technical contributions. (1) We offer simple, yet effective, feature aggregation methods, improving the prototypes used by ProtoNet, a popular FSL technique. (2) We describe a novel Transformer model for Noisy Few-Shot Learning (TraNFS). TraNFS leverages a transformer's attention mechanism to weigh mislabeled versus correct samples. (3) Finally, we extensively test these methods on noisy versions of MiniImageNet and TieredImageNet. Our results show that TraNFS is on-par with leading FSL methods on clean support sets, yet outperforms them, by far, in the presence of label noise.