The cascade-correlation learning architecture
Advances in neural information processing systems 2
Feature selection for pattern classification with Gaussian mixture models: a new objective criterion
Pattern Recognition Letters
| Hi-index | 0.00 |
This paper compares the performance of Generalised Fisher Ratio, Normalised Entropy and Functional Merging on assessing the exact parameters and number of components of finite Gaussian mixtures. The three methods use different statistical criteria for calculating the right number of components and have in common the Expectation-Maximisation algorithm for assessing the parameters of Gaussian mixtures. After applying the three above criteria to a similar benchmark as in paper (Celeux and Soromenho, Journal of Classification, 13, 1996) we show that Functional Merging surpass the performance of the other two methods.