Skip to content
Research Article Open access CC BY 4.0

On Designing Invertible Pseudo Covariance Matrix for Undersampled Cases in Classification

Rashid Mahmood, Khalid Mahmood Aamir, Marija Milojević Jevrić, Stojan Radenović, Tehseen Zia

Journal of Advances in Mathematics and Computer Science · pp. 1–9 · Published 9 Jul 2016

10.9734/BJMCS/2016/27435

Abstract

In linear discriminant analysis, determinant and inverse of the covariance matrix are required to be computed. If number of features is greater than the number of available examples, covariance matrix is no longer invertible. A common approach is to reduce dimensionality due to which some features of interest may be lost. When we are not interested in dimensionality reduction, one approach to solve such problems is to take pseudoinverse of covariance matrix which is not always possible. We propose, in such cases, to project covariance matrix onto a highly correlated space to compute pseudoinverse of the matrix. Proposed solution has been tested for classification of microarray gene expression data of colon’s tumor.

Covariance matrix dimensionality reduction pseudoinverse.

References (17)

  1. 1 Eigenfaces vs. Fisherfaces: recognition using class specific linear projection [DOI]
  2. 2 Using discriminant eigenfeatures for image retrieval [DOI]
  3. 3 A direct LDA algorithm for high-dimensional data — with application to face recognition [DOI]
  4. 4 Eigenfaces vs. Fisherfaces: Recognition using class specific linear projection [DOI]
  5. 5 Penalized Discriminant Analysis [DOI]
  6. 6 A Statistical Perspective on Ill-Posed Inverse Problems [DOI]
  7. 7 Why can LDA be performed in PCA transformed space? [DOI]
  8. 8 Generalizing discriminant analysis using the generalized singular value decomposition [DOI]
  9. 9 Expected classification error of the Fisher linear classifier with pseudo-inverse covariance matrix [DOI]
  10. 10 Common Structure of Smoothing Techniques in Statistics [DOI]
  11. 11 Discriminant Analysis with Singular Covariance Matrices: Methods and Applications to Spectroscopic Data [DOI]
  12. 12 An efficient algorithm to solve the small sample size problem for LDA [DOI]
  13. 13 Stabilizing classifiers for very small sample sizes [DOI]
  14. 14 Real-time face recognition using Gram-Schmidt orthogonalization for LDA [DOI]
  15. 15 Review of the International Statistical Institute [DOI]
  16. 16 Discriminant Functions [DOI]
  17. 17 Characterizations of H-J matrices [DOI]

Cited by 0

No indexed citations yet.

Related research

Article metrics

Real usage data collected on this platform.

30

Page views

0

PDF downloads

2

Outbound clicks

0

Citations

Views over time

Views by country

Approximate, from request IP at view time — not citizenship or institution. Countries with fewer than 5 views are grouped as "Other".

Traffic sources

Referring site, by host.

Views and downloads exclude known bots/crawlers. Citations combines this platform's own DOI-resolved index with each external source's own reported total — see Cited by above for individually listed citing works. Last refreshed 0 seconds ago.