Reproduce “Active learning for data streams: a survey” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Davide Cacciarelli, Murat Kulahci
This paper presents a comprehensive survey of active learning strategies tailored for continuous data streams, consolidating diverse query methods and addressing challenges unique to dynamic, high-velocity environments. It structures the field with a novel taxonomy and identifies critical areas for future research, particularly in regression and single-pass sampling.
Key insights
- Existing online active learning research has predominantly focused on classification tasks, with limited attention to online active linear or general regression models.
- Concept drift is a major challenge in data streams, requiring online active learning strategies to actively adapt through integration with drift detectors, ensemble methods, or adaptive budget allocation.
- Effective online active learning strategies are often dynamic, adaptive, and hybrid, employing diverse query methods (uncertainty, disagreement, diversity) and data processing paradigms (single-pass, window-based).
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.