File(s) stored somewhere else

Please note: Linked content is NOT stored on Carnegie Mellon University and we can't guarantee its availability, quality, security or accept any liability.

Dual Strategy Active Learning

journal contribution
posted on 01.01.2007 by Pinar Donmez, Jaime G. Carbonell, Paul N. Bennett

Active Learning methods rely on static strategies for sampling unlabeled point(s). These strategies range from uncertainty sampling and density estimation to multi-factor methods with learn-once-use-always model parameters. This paper proposes a dynamic approach, called DUAL, where the strategy selection parameters are adaptively updated based on estimated future residual error reduction after each actively sampled point. The ob jective of dual is to outperform static strategies over a large operating range: from very few to very many labeled points. Empirical results over six datasets demonstrate that DUAL outperforms several state-of-the-art methods on most datasets.