TiDAL: Learning Training Dynamics for Active Learning

TL;DR

TiDAL introduces a novel active learning approach that leverages training dynamics to better estimate data uncertainty, leading to improved sample selection and model performance over traditional static uncertainty methods.

Contribution

The paper proposes TiDAL, a new active learning method that incorporates training dynamics via a learned prediction module, providing a more effective uncertainty measure.

Findings

TiDAL outperforms existing methods on benchmark datasets.

Training dynamics provide valuable uncertainty information.

TiDAL is effective for both balanced and imbalanced datasets.

Abstract

Active learning (AL) aims to select the most useful data samples from an unlabeled data pool and annotate them to expand the labeled dataset under a limited budget. Especially, uncertainty-based methods choose the most uncertain samples, which are known to be effective in improving model performance. However, AL literature often overlooks training dynamics (TD), defined as the ever-changing model behavior during optimization via stochastic gradient descent, even though other areas of literature have empirically shown that TD provides important clues for measuring the sample uncertainty. In this paper, we propose a novel AL method, Training Dynamics for Active Learning (TiDAL), which leverages the TD to quantify uncertainties of unlabeled data. Since tracking the TD of all the large-scale unlabeled data is impractical, TiDAL utilizes an additional prediction module that learns the TD of labeled data. To further justify the design of TiDAL, we provide theoretical and empirical evidence to argue the usefulness of leveraging TD for AL. Experimental results show that our TiDAL achieves better or comparable performance on both balanced and imbalanced benchmark datasets compared to state-of-the-art AL methods, which estimate data uncertainty using only static information after model training.