Rethinking deep active learning: Using unlabeled data at model training

TL;DR

This paper proposes a novel active learning approach that incorporates unlabeled data during training through unsupervised and semi-supervised methods, leading to significant accuracy improvements in image classification.

Contribution

It introduces the use of unlabeled data during training in active learning, combining unsupervised feature learning and semi-supervised learning, which has not been extensively studied before.

Findings

Unlabeled data during training improves accuracy significantly.

Semi-supervised learning benefits are consistent across datasets.

Effective even with minimal labels, such as one per class.

Abstract

Active learning typically focuses on training a model on few labeled examples alone, while unlabeled ones are only used for acquisition. In this work we depart from this setting by using both labeled and unlabeled data during model training across active learning cycles. We do so by using unsupervised feature learning at the beginning of the active learning pipeline and semi-supervised learning at every active learning cycle, on all available data. The former has not been investigated before in active learning, while the study of latter in the context of deep learning is scarce and recent findings are not conclusive with respect to its benefit. Our idea is orthogonal to acquisition strategies by using more data, much like ensemble methods use more models. By systematically evaluating on a number of popular acquisition strategies and datasets, we find that the use of unlabeled data during model training brings a surprising accuracy improvement in image classification, compared to the differences between acquisition strategies. We thus explore smaller label budgets, even one label per class.