Learning to Enumerate

Conference paper
Part of the Lecture Notes in Computer Science book series (LNCS, volume 9886)

Abstract

The Learning to Enumerate problem is a new variant of the typical active learning problem. Our objective is to find data that satisfies arbitrary but fixed conditions, without using any prelabeled training data. The key aspect here is to query as few as possible non-target data. While typical active learning techniques try to keep the number of queried labels low they give no regards to the class these instances belong to. Since the aim of this problem is different from the common active learning problem, we started with applying uncertainty sampling as a base technique and evaluated the performance of three different base learner on 19 public datasets from the UCI Machine Learning Repository.

Keywords

Active learning Learning to enumerate Exploration vs. exploitation Epsilon-greedy 

References

  1. 1.
    Pedregosa, F., Varoquaux, G., Gramfort, A., Michel, V., Thirion, B., Grisel, O., Blondel, M., Prettenhofer, P., Weiss, R., Dubourg, V., Vanderplas, J., Passos, A., Cournapeau, D., Brucher, M., Perrot, M., Duchesnay, E.: Scikit-learn: machine learning in Python. J. Mach. Learn. Res. 12, 2825–2830 (2011)MathSciNetMATHGoogle Scholar
  2. 2.
    Settles, B.: Active Learning. Synth. Lect. Artif. Intell. Mach. Learn. 6, 1–114 (2012). Morgan & Claypool PublishersMathSciNetCrossRefMATHGoogle Scholar
  3. 3.
    Baba, Y., Kashima, H., Nohara, Y., Kai, E., Ghosh, P., Islam, R., Ahmed, A., Kuruda, M., Inoue, S., Hiramatsu, T., Kimura, M., Shimizu, S., Kobayashi, K., Tsuda, K., Sugiyama, M., Blondel, M., Ueda, N., Kitsuregawa, M., Nakashima, N.: Predictive approaches for low-cost preventive medicine program in developing countries. In: Proceedings of the 21st ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, pp. 1681–1690. ACM (2015)Google Scholar
  4. 4.
    Kajino, H., Kishimoto, A., Botea, A., Daly, E., Kotoulas, S.: Active learning for multi-relational data construction. In: Proceedings of the 24th International Conference on World Wide Web, pp. 560–569. ACM (2015)Google Scholar
  5. 5.
    Auer, P., Cesa-Bianchi, N., Fischer, P.: Finite-time analysis of the multiarmed bandit problem. Mach. Learn. 47(2–3), 235–256 (2002). SpringerCrossRefMATHGoogle Scholar
  6. 6.
  7. 7.
    UC Irvine Machine Learning Repository. https://archive.ics.uci.edu/ml/index.html

Copyright information

© Springer International Publishing Switzerland 2016

Authors and Affiliations

  • Patrick Jörger
    • 1
    • 2
  • Yukino Baba
    • 1
  • Hisashi Kashima
    • 1
  1. 1.Kyoto UniversityKyotoJapan
  2. 2.Karlsruhe Institute of TechnologyKarlsruheGermany

Personalised recommendations