Optimal use of a trained neural network for input selection
In this paper, we present a review of feature selection methods, based on the analysis of a trained multilayer feedforward network, which have been applied to neural networks. Furthermore, a methodology that allows evaluating and comparing feature selection methods is carefully described. This methodology is applied to the 19 reviewed methods in a total of 15 different real world classification problems. We present an ordination of methods according to its performance and it is clearly concluded which method performs better and should be used. We also discuss the applicability and computational complexity of the methods.
Unable to display preview. Download preview PDF.
- 1.Devena, L.: Automatic selection of the most relevant features to recognize objects. Proc. of the Int. Conf. on Artificial NNs, vol. 2, pp. 1113–1116, 1994.Google Scholar
- 4.Engelbrecht, AP., Cloete, I.: A sensitivity analysis algorithm for pruning feedforward neural networks. Proc. of the Int. Conf. on Neural Networks, vol. 2, pp. 1274–1277, 1996.Google Scholar
- 6.Lee, H., Mehrotra, K., Mohan, C. Ranka, S.: Selection procedures for redundant inputs in neural networks. Proc. of the World Congress on Neural Networks, vol. 1, pp. 300–303, 1993.Google Scholar
- 7.Tetko, I.V., Tanchuk, V.Y., Luik, A.I.: Simple heuristic methods for input parameter estimation in neural networks. Proc. of the IEEE Int. Conf. on Neural Networks, vol. 1, pp. 376–380, 1994.Google Scholar
- 10.El-Deredy, W., Branston, N.M.: Identification of relevant features in HMR tumor spectra using neural networks. Proc. of the 4th Int. Conf. on Artificial Neural Networks, pp. 454–458, 1995.Google Scholar
- 12.Mao, J., Mohiuddin, K., Jain, A.K.: Parsimonious network design and feature selection through node pruning. Proc. of the 12th IAPR Int. Conf. on Pattern Recognition, vol. 2, pp. 622–624, 1994.Google Scholar
- 13.Utans, J., Moody, J., Rehfuss, S., Siegelmann, H.: Input variable selection for neural networks: Application to predicting the U.S. business cycle. Proc. of IEEE/IAFE 1995 Comput. Intellig. for Financial Eng., pp. 118–122, 1995.Google Scholar
- 14.Younes, B., Fabrice, B.: A neural network based variable selector. Proc. of the Artificial Neural Network in Engineering, (ANNIE'95), pp. 425–430, 1995.Google Scholar
- 15.Bowles, A.: Machine learns which features to select”. Proc. of the 5th Australian Joint Conf. on Artificial Intelligence, pp. 127–132, 1992.Google Scholar
- 16.Sano, H., Nada, A., Iwahori, Y., Ishii, N.: A method of analyzing information represented in neural networks. Proc. of 1993 Int. Joint Conf. on Neural Networks, pp. 2719–2722, 1993.Google Scholar
- 17.Bishop, C.: Exact calculation of the hessian matrix for the multilayer perceptron. Neural Computation, vol. 4, pp. 494–501, 1992.Google Scholar
- 18.Watzel, R., Meyer-Bäse, A., Meyer-Bäse, U., hilberg, H., Scheich, H.: Identification of irrelevant features in phoneme recognition with radial basis classifiers. Proc. of 1994 Int. Symp. on Artificial NNs, pp. 507–512, 1994.Google Scholar
- 19.Bronshtein, I., Semandiavev, K.: Mathematics Handbook for engineers and students (in Spanish). MIR, Moscow, 1977.Google Scholar