Iterative label cleaning for transductive and semi-supervised few-shot learning

TL;DR

This paper introduces an iterative label cleaning algorithm for transductive and semi-supervised few-shot learning, leveraging data manifold structures to improve pseudo-label quality and achieve state-of-the-art results on multiple benchmarks.

Contribution

The paper presents a novel iterative label cleaning method that enhances pseudo-label accuracy by using manifold structures and loss distribution, improving few-shot learning performance.

Findings

Achieves state-of-the-art results on miniImageNet, tieredImageNet, CUB, and CIFAR-FS.

Robust across different feature pre-processing methods.

Effective with varying amounts of unlabeled data.

Abstract

Few-shot learning amounts to learning representations and acquiring knowledge such that novel tasks may be solved with both supervision and data being limited. Improved performance is possible by transductive inference, where the entire test set is available concurrently, and semi-supervised learning, where more unlabeled data is available. Focusing on these two settings, we introduce a new algorithm that leverages the manifold structure of the labeled and unlabeled data distribution to predict pseudo-labels, while balancing over classes and using the loss value distribution of a limited-capacity classifier to select the cleanest labels, iteratively improving the quality of pseudo-labels. Our solution surpasses or matches the state of the art results on four benchmark datasets, namely miniImageNet, tieredImageNet, CUB and CIFAR-FS, while being robust over feature space pre-processing and the quantity of available data. The publicly available source code can be found in https://github.com/MichalisLazarou/iLPC.