Feature selection for pattern classification with Gaussian mixture models: A new objective criterion

Krishnan, S.; Samudravijaya, K.; Rao, P. V.

doi:10.1016/0167-8655(96)00047-5

Cited by 25 publications

(11 citation statements)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…One reason for this could be that the center of gravity measurements obtained by pair wise Fischer discriminants is inadequate. Some alternative objective measures could be tried such as given by Lee [23] and Krishnan et al [20].…”

Section: Resultsmentioning

confidence: 99%

A Comparative Study of Explicit Frequency and Conventional Signal Representations for Speech Recognition

Krishnan

Rao

1996

Digital Signal Processing

Self Cite

View full text Add to dashboard Cite

Section: Resultsmentioning

confidence: 99%

A Comparative Study of Explicit Frequency and Conventional Signal Representations for Speech Recognition

Krishnan

Rao

1996

Digital Signal Processing

Self Cite

View full text Add to dashboard Cite

“…The best features in descending order of the Fisherratios can then be selected for the classification task [60,61]. In addition, the Fisher criterion can be extended to multi-class problems enabling the simultaneous distinction between several groups [61,62].…”

Section: Methodsmentioning

confidence: 99%

Thermal desorption/pyrolysis coupled with photo ionisation time-of-flight mass spectrometry for the analysis and discrimination of pure tobacco samples

Streibel

Mitschke

Adam

et al. 2007

Journal of Analytical and Applied Pyrolysis

View full text Add to dashboard Cite

“…While a number of previous studies that investigated feature selection performance also used simulated Gaussian data, 2,8,9,14 a few others simulated a mixture of Gaussians 15,16 and Boolean feature spaces. 17,18 Although clinical data may not follow any of these idealized distributions, we chose to use Gaussian distributions because they are commonly used in both simulation studies and theoretical analyses of classifier performance in pattern recognition literature.…”

Section: Iia Class Distributionsmentioning

confidence: 99%

Effect of finite sample size on feature selection and classification: A simulation study

et al. 2010

View full text Add to dashboard Cite

Purpose:The small number of samples available for training and testing is often the limiting factor in finding the most effective features and designing an optimal computer-aided diagnosis ͑CAD͒ system. Training on a limited set of samples introduces bias and variance in the performance of a CAD system relative to that trained with an infinite sample size. In this work, the authors conducted a simulation study to evaluate the performances of various combinations of classifiers and feature selection techniques and their dependence on the class distribution, dimensionality, and the training sample size. The understanding of these relationships will facilitate development of effective CAD systems under the constraint of limited available samples. Methods: Three feature selection techniques, the stepwise feature selection ͑SFS͒, sequential floating forward search ͑SFFS͒, and principal component analysis ͑PCA͒, and two commonly used classifiers, Fisher's linear discriminant analysis ͑LDA͒ and support vector machine ͑SVM͒, were investigated. Samples were drawn from multidimensional feature spaces of multivariate Gaussian distributions with equal or unequal covariance matrices and unequal means, and with equal covariance matrices and unequal means estimated from a clinical data set. Classifier performance was quantified by the area under the receiver operating characteristic curve A z . The mean A z values obtained by resubstitution and hold-out methods were evaluated for training sample sizes ranging from 15 to 100 per class. The number of simulated features available for selection was chosen to be 50, 100, and 200. Results: It was found that the relative performance of the different combinations of classifier and feature selection method depends on the feature space distributions, the dimensionality, and the available training sample sizes. The LDA and SVM with radial kernel performed similarly for most of the conditions evaluated in this study, although the SVM classifier showed a slightly higher hold-out performance than LDA for some conditions and vice versa for other conditions. PCA was comparable to or better than SFS and SFFS for LDA at small samples sizes, but inferior for SVM with polynomial kernel. For the class distributions simulated from clinical data, PCA did not show advantages over the other two feature selection methods. Under this condition, the SVM with radial kernel performed better than the LDA when few training samples were available, while LDA performed better when a large number of training samples were available. Conclusions: None of the investigated feature selection-classifier combinations provided consistently superior performance under the studied conditions for different sample sizes and feature space distributions. In general, the SFFS method was comparable to the SFS method while PCA may have an advantage for Gaussian feature spaces with unequal covariance matrices. The performance of the SVM with radial kernel was better than, or comparable to, that of the SVM with polynomial kernel un...

show abstract

Feature selection for pattern classification with Gaussian mixture models: A new objective criterion

Cited by 25 publications

References 10 publications

A Comparative Study of Explicit Frequency and Conventional Signal Representations for Speech Recognition

A Comparative Study of Explicit Frequency and Conventional Signal Representations for Speech Recognition

Thermal desorption/pyrolysis coupled with photo ionisation time-of-flight mass spectrometry for the analysis and discrimination of pure tobacco samples

Effect of finite sample size on feature selection and classification: A simulation study

Contact Info

Product

Resources

About