• Summary
• Related Documents
  • Active Bibliography
  • Co-citation
• Version History

Abstract:

multilay er perceptron isusually considered a passive learner thatonly receives given training data. However, if a multilay er perceptron actively gathers training data that resolve its uncertainty about a problem being learnt, su#ciently accurate classification is attained with fewer training data. Recently, such active learning has been receiving an increasing interest. In this paper, we propose a novel active learningstrategy hestrategy attempts to produceonly useful training data for multilay er perceptrons to achieve accurate classification, and avoids generating redundant training data. Furthermore, thestrategy attempts to avoid generating temporarily useful training data that will become redundant in the future. As a result, thestrategy can allow multilay er perceptrons to achieve accurate classification with fewer training data. o demonstrate the performance of the strategy in comparison with other active learning strategies, we also propose an empirical active learning algorithm as an implementation of thestrategy, which does not require expensive computations. Experimental results show that the proposed algorithm improves the classificationaccuracy of a multilay er perceptron with fewer training data than that for a conventional random selection algorithm that constructs a training data set without explicit strategies. Moreover, the algorithm outperforms ty/6 cal active learning algorithms in the experiments. hose results show that the algorithm can construct an appropriate training data set at lower computational cost, because training data generation isusually costly

Citations