作者: D.H. Foley , J.W Sammon
关键词:
摘要: A new method for the extraction of features in a two-class pattern recognition problem is derived. The main advantage that selecting based entirely upon discrimination or separability as opposed to more common approach fitting. classical example fitting use eigenvectors lumped covariance matrix corresponding largest eigenvalues. In an analogous manner, technique selects discriminant vectors (or features) "discrim-values." compared some popular alternative techniques via both data-dependent and mathematical examples. addition, recursive obtaining given.