Depth and Width Changeable Network-Based Deep Kernel Learning-Based Hyperspectral Sensor Data Analysis

Table 5

The classification performance by comparing with existing MKL algorithm.

Dataset	EasyMKL^1[22]	SimpleMKL^2[23]	SM1MKL^3[24]	L2MKL^4[25, 26]	Our method (D,W)
Australian	84.63 (3)	84.30 (4)	85.79 (1)	84.05 (5)	85.50 (3,3) (2)
Breast	69.06 (5)	69.78 (4)	70.50 (1)	70.50 (1)	70.50 (1,1) (1)
Cleve	81.08 (4)	80.40 (5)	81.75 (2)	81.75 (2)	84.45 (3,3) (1)
Cloud	69.78 (4)	75.43 (2)	73.29 (3)	61.20 (5)	99.80 (1,3) (1)
Diabetes	74.21 (5)	74.73 (4)	76.56 (2)	76.56 (2)	77.34 (3,3) (1)
Pima	78.38 (1)	78.12 (3)	77.34 (4)	77.34 (4)	78.38 (3,3) (1)
Sonar	86.66 (1)	76.69 (3)	72.38 (5)	74.28 (4)	84.76 (1,3) (2)
Average rank	3.27	2.14	2.57	3.29	1.28

¹Avaliable at https://github.com/okbalefthanded/EasyMKL. ²Avaliable at https://github.com/maxis1718/SimpleMKL. ³Avaliable at https://sites.google.com/site/xinxingxu666/. ⁴The implementation is obtained in footnote 4.