Using Unlabeled Data for Learning Classification Problems
This chapter presents an approach of using unlabeled data for learning classification problems. The chapter consists of two parts. In the first part of the chapter, an approach of using both labeled and unlabeled data to train a multilayer perceptron is presented. The approach banks on the assumption that regions of low pattern density usually separate data classes. The unlabeled data are iteratively preprocessed by a perceptron being trained to obtain the soft class label estimates. It is demonstrated that substantial gains in classification performance may be achieved by using the approach when the labeled data do not adequately represent the entire class distributions. In the second part of the chapter, we propose a quality function for learning decision boundary between data clusters from unlabeled data. The function is based on third order polynomials. The objective of the quality function is to find a place in the input space sparse in data points. By maximizing the quality function, we find a decision boundary between data clusters. A superiority of the proposed quality function over the other similar functions as well as the conventional clustering algorithms tested has been observed in the experiments.
KeywordsQuality Function Decision Boundary Multilayer Perceptron Label Data Unlabeled Data
Unable to display preview. Download preview PDF.
- Cataltepe, Z. and Magdon-Ismail, M. (1998), “Incorporating test inputs into learning,” Jordan, M.I., Kearns, M.J., and Solla, S.A. (Eds.), Advances in Neural Information Processing Systems 10, MIT Press, pp. 437–443.Google Scholar
- Ghosh, J. and Shin, Y. (1992), “Efficient higher-order neural networks for classification and function approximation,” International Journal of Neural Systems, pp. 323–350.Google Scholar
- Hyvärinen, A. and Oja, E. (1997), “One-unit learning rules for independent component analysis,” Mozer, M.C., Jordan, M.I., and Petsche, T. (Eds.), Advances in Neural Information Processing Systems 9, MIT Press, pp. 480–486.Google Scholar
- Kosko, B. (1992), Neural Networks and Fuzzy Systems. A Dynamical Systems Approach to Machine Intelligence,Prentice-Hall.Google Scholar
- Lippman, R.P. (1987), “An introduction to computing with neural nets,” IEEE ASSP Magazine, pp. 4–22.Google Scholar
- Miller, D.J. and Uyar, H.S. (1997), “A mixture of experts classifier with learning based on both labeled and unlabeled data,” Mozer, M.C., Jordan, M.I., and Petsche, T. (Eds.), Advances in Neural Information Processing Systems 9, MIT Press, pp. 571–577.Google Scholar
- Osterberg, M. (1994), Quality functions for parallel selective principal component analysis, Ph.D. Thesis, Linköping University, UniTryck, Linköping.Google Scholar
- Towell, G. (1997), “Using unlabeled data for supervised learning,” Mozer, M.C., Jordan, M.I., and Petsche, T. (Eds.), Advances in Neural Information Processing Systems 9, MIT Press, pp. 647–653.Google Scholar
- Tsao, C.E., Bezdek, J.C., and Pal, N.R. (1996), “Fuzzy Kohonen clustering networks,” Pattern Recognition, vol. 29, pp. 757–764.Google Scholar
- Verikas, A., Gelzinis, A., and Malmqvist, K. (1997), “A random search technique for training neural networks,” Proceedings of the fourth International Conference on Neural Information Processing, ICONIP’97, Dunedin, pp. 322–325.Google Scholar
- Waxman, A.M., Seibert, M., Gove, A.N., Fay, D.A., Cunningham, R.K., and Bachelder, I.A. (1994), “Visual learning of objects: Neural models of shape, color, motion and space,” Zurada, J.M., Marks, R.J., and Robinson, C.J. (Eds.), Computational Intelligence Imitating Life, IEEE Press, pp. 237–251.Google Scholar