Table 3.

Comparing performance of five kernel-based dimensionality reduction algorithms

Uni-Biometric Iris Recognition Right-CASIA Dim reduction algorithm|Kernel Function	Future extraction: Daugman Algorithm. Classification: Dis-Angle (%)								Future Extraction: Hough Tr. Classification: Dis-Angle (%) Dimensionality of feature space Dimensionality of feature space
	3	5	15	35	50	80	100	150	3	5	15	35	50	80	100	150
Kernel LDA
Gaussian	75.93	86.11	90.74	93.52	93.52	92.59	92.59	93.52	75	83.33	87.96	93.52	94.44	97.22	95.37	95.3
PolyPlus	14.81	40.74	78.70	90.74	93.52	93.52	94.44	94.44	12.92	26.85	78.70	87.96	89.81	95.37	97.22	97.22
Polynomial	12.4	38.89	81.48	94.44	94.44	94.44	95.37	94.44	<10	31.48	73.15	84.26	92.60	95.37	96.30	97.22
Linear	11.11	25.93	78.70	92.59	92.59	94.44	94.44	94.44	<10	25.93	68.52	87.04	87.04	92.59	95.37	97.22
Hamming	<10	24.07	76.85	87.96	91.67	92.59	94.44	94.44	<10	234.07	76.85	86.11	86.11	95.37	96.30	97.22
Kernel PCA
Gaussian	<10	<10	14.81	27.78	37.04	53.70	64.81	78.70	<10	<10	26.85	42.60	50	62.56	69.44	82.42
PolyPlus	<10	21.30	71.30	90.74	92.59	96.30	96.30	95.37	<10	21.30	62.56	87.90	91.67	94.44	93.52	93.52
Polynomial	<10	21.30	71.30	90.74	92.59	96.30	96.30	95.37	<10	21.30	62.56	87.90	91.67	94.44	93.52	93.52
Linear	<10	22.22	69.44	89.81	92.59	95.37	95.37	94.44	<10	21.30	67.60	87.90	92.59	95.37	94.44	96.30
None
Dim-9600	93.52								93.52

LDA – Linear discriminant analysis; PCA – Principal component analysis