Extraction and Recognition of Nonlinear Interval-Type Features Using Symbolic KDA Algorithm with Application to Face Recognition

Table 1

Optimal parameters corresponding to each method with respect to two different kernels.

Method	Polynomial kernel		Gaussian kernel
	Order	Subspace dimension	Width	Subspace dimension
Kernel Eigenfaces	1	44	7	47
Kernel fisherfaces	3	35	5	44
Symbolic KDA	3	18	4	24