An Active Learning Approach for Training the Probabilistic RBF Classification Network

Constantinos Constantinopoulos, Aristidis Likas

Abstract

Active learning for classification constitutes a type of learning problem where a classifier is gradually built by iteratively asking for the labels of data points. The method involves a data selection mechanism that queries for the labels of those data points that considers to be mostly beneficial for improving the performance of the current classifier. We present an active learning methodology for training the probabilistic RBF (PRBF) network which is a special case of the RBF network, and constitutes a generalization of the Gaussian mixture model. The method employs a suitable criterion to select an unlabeled observation and query its label. The proposed criterion selects points that lie near the decision boundary. The learning performance of the algorithm is tested with experiments on several data sets.

References

  1. Bishop, C. M. (1995). Neural Networks for Pattern Recognition. Oxford University Press.
  2. Castelli, V. and Cover, T. (1995). On the exponential value of labeled samples. Pattern Recognition Letters, 16:105-111.
  3. Cohn, D., Ghahramani, Z., and Jordan, M. (1996). Active learning with statistical models. Journal of Artificial Inteligence Research, 4:129-145.
  4. Constantinopoulos, C. and Likas, A. (2006). An incremental training method for the probabilistic RBF network. IEEE Trans. Neural Networks, to appear.
  5. Dempster, A., Laird, N., and Rubin, D. (1977). Maximum likelihood estimation from incomplete data via the EM algorithm. Journal of the Royal Statistical Society, Series B, 39(1):1-38.
  6. Freund, Y., Seung, H. S., Shamir, E., and Tishby, N. (1997). Selective sampling using the query by comitee algorithm. Machine Learning, 28:133-168.
  7. McCallum, A. K. and Nigam, K. (1998). Employing EM in pool-based active learning for text classification. In Shavlik, J. W., editor, Proc. 15th International Conference on Machine Learning. Morgan Kaufmann.
  8. McLachlan, G. and Krishnan, T. (1997). The EM algorithm and extensions. John Wiley & Sons.
  9. McLachlan, G. and Peel, D. (2000). Finite Mixture Models. John Wiley & Sons.
  10. Titsias, M. K. and Likas, A. (2000). A probabilistic RBF network for classification. In Proc. International Joint Conference on Neural Networks 4, pages 238-243. IEEE.
  11. Titsias, M. K. and Likas, A. (2001). Shared kernel models for class conditional density estimation. IEEE Trans. Neural Networks, 12(5):987-997.
  12. Titsias, M. K. and Likas, A. (2002). Mixture of experts classification using a hierarchical mixture model. Neural Computation, 14(9):2221-2244.
  13. Titsias, M. K. and Likas, A. (2003). Class conditional density estimation using mixtures with constrained component sharing. IEEE Trans. Pattern Anal. and Machine Intell., 25(7):924- 928.
  14. Zhu, X., Lafferty, J., and Ghahramani, Z. (2003). Combining active learning and semisupervised learning using Gaussian fields and harmonic functions. In Proc. 20th International Conference on Machine Learning.
Download


Paper Citation


in Harvard Style

Constantinopoulos C. and Likas A. (2006). An Active Learning Approach for Training the Probabilistic RBF Classification Network . In Proceedings of the 2nd International Workshop on Artificial Neural Networks and Intelligent Information Processing - Volume 1: ANNIIP, (ICINCO 2006) ISBN 978-972-8865-68-9, pages 3-12. DOI: 10.5220/0001222800030012


in Bibtex Style

@conference{anniip06,
author={Constantinos Constantinopoulos and Aristidis Likas},
title={An Active Learning Approach for Training the Probabilistic RBF Classification Network},
booktitle={Proceedings of the 2nd International Workshop on Artificial Neural Networks and Intelligent Information Processing - Volume 1: ANNIIP, (ICINCO 2006)},
year={2006},
pages={3-12},
publisher={SciTePress},
organization={INSTICC},
doi={10.5220/0001222800030012},
isbn={978-972-8865-68-9},
}


in EndNote Style

TY - CONF
JO - Proceedings of the 2nd International Workshop on Artificial Neural Networks and Intelligent Information Processing - Volume 1: ANNIIP, (ICINCO 2006)
TI - An Active Learning Approach for Training the Probabilistic RBF Classification Network
SN - 978-972-8865-68-9
AU - Constantinopoulos C.
AU - Likas A.
PY - 2006
SP - 3
EP - 12
DO - 10.5220/0001222800030012