The Theory Analysis on FSDVS and an Optimal Model
Yang Jing · 2003
The attribute of the maximum of R(ξ) in arbitrary subspace of Rn is discussed dedicatedly. The theory analysis indicates that every F.S discriminant vector is better than respective vector in other discriminant vectors sets, which consist of eigenvectors of sbζ = λSwξ. But,the fact that the F.S vectors are statistically correlated degrade the F. S vectors set. Two ways are used to obtaingood F-S vectors set. The experiments on Concordia University CEN-PARMI handwritten numeral database suggest that the new vectors set is better than the origin. A new problem model on F-S discriminant vectors set is proposed in this paper. It's easily understanding that the problem model is superior to the original F-S discriminant vectors set problem model.