Unsupervised Few-shot Learning via Deep Laplacian Eigenmaps

TL;DR

This paper introduces an unsupervised few-shot learning approach using deep Laplacian eigenmaps, enabling effective representation learning from unlabeled data and bridging the gap with supervised methods.

Contribution

It proposes a novel unsupervised learning method that leverages deep Laplacian eigenmaps for few-shot tasks, avoiding collapsed representations without labeled data.

Findings

Achieves performance close to supervised few-shot learning.

Comparable to state-of-the-art self-supervised methods.

Effectively groups similar samples in unlabeled data.

Abstract

Learning a new task from a handful of examples remains an open challenge in machine learning. Despite the recent progress in few-shot learning, most methods rely on supervised pretraining or meta-learning on labeled meta-training data and cannot be applied to the case where the pretraining data is unlabeled. In this study, we present an unsupervised few-shot learning method via deep Laplacian eigenmaps. Our method learns representation from unlabeled data by grouping similar samples together and can be intuitively interpreted by random walks on augmented training data. We analytically show how deep Laplacian eigenmaps avoid collapsed representation in unsupervised learning without explicit comparison between positive and negative samples. The proposed method significantly closes the performance gap between supervised and unsupervised few-shot learning. Our method also achieves comparable performance to current state-of-the-art self-supervised learning methods under linear evaluation protocol.