Towards Computationally Feasible Deep Active Learning

TL;DR

This paper introduces techniques to make deep active learning more computationally feasible for text classification by reducing training time and overhead, while also improving model performance through pseudo-labeling and model distillation.

Contribution

The authors propose two methods to reduce computational costs in deep active learning and demonstrate that pseudo-labeling with distilled models enhances success despite using smaller acquisition models.

Findings

Significant reduction in active learning iteration duration.

Improved model performance with pseudo-labeling and distillation.

Overcoming the gap between acquisition and successor models.

Abstract

Active learning (AL) is a prominent technique for reducing the annotation effort required for training machine learning models. Deep learning offers a solution for several essential obstacles to deploying AL in practice but introduces many others. One of such problems is the excessive computational resources required to train an acquisition model and estimate its uncertainty on instances in the unlabeled pool. We propose two techniques that tackle this issue for text classification and tagging tasks, offering a substantial reduction of AL iteration duration and the computational overhead introduced by deep acquisition models in AL. We also demonstrate that our algorithm that leverages pseudo-labeling and distilled models overcomes one of the essential obstacles revealed previously in the literature. Namely, it was shown that due to differences between an acquisition model used to select instances during AL and a successor model trained on the labeled data, the benefits of AL can diminish. We show that our algorithm, despite using a smaller and faster acquisition model, is capable of training a more expressive successor model with higher performance.