Covariance Matrix Estimation and Classification With Limited Training Data
|
Abstract—A new covariance matrix estimator useful for designing classifiers with limited training data is developed. In experiments, this estimator achieved higher classification accuracy than the sample covariance matrix and common covariance matrix estimates. In about half of the experiments, it achieved higher accuracy than regularized discriminant analysis, but required much less computation.
[1] 763 H.W. Sorenson, Parameter Estimation: Principles and Problems, pp. 183-184.New York: M. Dekker, 1980.
[2] T.W. Anderson, An Introduction to Multivariate Statistical Analysis, second edition, p. 209.New York: John Wiley&Sons, 1984.
[3] J.H. Friedman, "Regularized Discriminant Analysis," J. Am. Statistical Assoc., vol. 84, pp. 165-175, Mar. 1989.
[4] K. Fukunaga, Introduction to Statistical Pattern Recognition, second edition. Academic Press, 1990.
[5] G.H. Golub and C.F. Van Loan, Matrix Computations, second edition, p. 51.Baltimore: Johns Hopkins Univ. Press, 1989.
[6] J.P. Hoffbeck, "Classification of High Dimensional Multispectral Data," PhD thesis, Purdue Univ., West Lafayette, Ind., pp. 55-70.
Index Terms:
Covariance matrix, estimation, leave-one-out method, cross validation, classification, high dimensional data.
Citation:
Joseph P. Hoffbeck, David A. Landgrebe, "Covariance Matrix Estimation and Classification With Limited Training Data," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 18, no. 7, pp. 763-767, July 1996, doi:10.1109/34.506799
PURCHASE ARTICLE: $19
IEEE Xplore Subscribers
Digg
Furl
Spurl
Blink
Simpy
Google
Del.icio.us
Y!MyWeb