Semi-Supervised Learning of Visual Features by Non-Parametrically Predicting View Assignments with Support Samples

TL;DR

PAWS introduces a semi-supervised learning method that non-parametrically predicts view assignments using support samples, achieving state-of-the-art results with less training on ImageNet.

Contribution

The paper presents a simple yet effective semi-supervised learning approach that extends distance-metric loss with non-parametric pseudo-labeling using support samples.

Findings

Outperforms previous semi-supervised methods on ImageNet.

Achieves 75.5% top-1 accuracy with 10% labels.

Requires significantly less training time.

Abstract

This paper proposes a novel method of learning by predicting view assignments with support samples (PAWS). The method trains a model to minimize a consistency loss, which ensures that different views of the same unlabeled instance are assigned similar pseudo-labels. The pseudo-labels are generated non-parametrically, by comparing the representations of the image views to those of a set of randomly sampled labeled images. The distance between the view representations and labeled representations is used to provide a weighting over class labels, which we interpret as a soft pseudo-label. By non-parametrically incorporating labeled samples in this way, PAWS extends the distance-metric loss used in self-supervised methods such as BYOL and SwAV to the semi-supervised setting. Despite the simplicity of the approach, PAWS outperforms other semi-supervised methods across architectures, setting a new state-of-the-art for a ResNet-50 on ImageNet trained with either 10% or 1% of the labels, reaching 75.5% and 66.5% top-1 respectively. PAWS requires 4x to 12x less training than the previous best methods.