Active learning (AL) traditionally relies on some instance-based utility measures (such as uncertainty) to assessindividual instances and label the ones with the maximum values for training. In this paper, we argue that such approaches cannotproduce good labeling subsets mainly because instances are evaluated independently without considering their interactions, and in-dividuals with maximal ability do not necessarily form an optimal instance subset for learning. Alternatively, we propose to achieveAL with optimal subset selection (ALOSS), where the key is to find an instance subset with a maximum utility value. To achievethe goal, ALOSS simultaneously considers the following: 1) the importance of individual instances and 2) the disparity betweeninstances, to build an instance-correlation matrix. As a result, AL is transformed to a semidefinite programming problem to selecta k -instance subset with a maximum utility value. Experimental results demonstrate that ALOSS outperforms state-of-the-artapproaches for AL.
|Number of pages||12|
|Journal||IEEE Transactions on Systems, Man, and Cybernetics, Part B: Cybernetics|
|Publication status||Published - Apr 2013|