Active Learning with Task-Driven Representations for Messy Pools
This addresses the challenge of selecting relevant data in messy pools for active learning, offering a novel method that is incremental but impactful for specific applications.
The paper tackles the problem of active learning in messy, uncurated data pools by proposing task-driven representations updated during the process, which significantly improve empirical performance over using fixed unsupervised representations.
Active learning has the potential to be especially useful for messy, uncurated pools where datapoints vary in relevance to the target task. However, state-of-the-art approaches to this problem currently rely on using fixed, unsupervised representations of the pool, focusing on modifying the acquisition function instead. We show that this model setup can undermine their effectiveness at dealing with messy pools, as such representations can fail to capture important information relevant to the task. To address this, we propose using task-driven representations that are periodically updated during the active learning process using the previously collected labels. We introduce two specific strategies for learning these representations, one based on directly learning semi-supervised representations and the other based on supervised fine-tuning of an initial unsupervised representation. We find that both significantly improve empirical performance over using unsupervised or pretrained representations.