Support Vector Machines for Classificationof Multi- and Hyperspectral Data
Different Optimizers using Linear Kernel
100
90
80
>u
70
I!!
::I
60
u
u
50
c(
'iii 40
... GI 30
>
0
20
10
0
c"
c"
c"
<:)"
" "
,,<:)
<:)<:)
<§><:)
c<:)
s§><:)
<§>"
<:)"
<:)' il
<:)0
<:)0
"
"r§S
"
<:)"
<:)0
<:)0
"
Penalty Value (C)
Fig. 10.7. Overall accuracy of different optimizers for the hyperspectrai image
Different Optimizers using Linear Kernel
3500 r.======::::;-----------r:1-~
3000 ___ CD - - L S V M f - - - - - - - - - - - - + - - T - - - - i
~2500~==========~----------_n~~~~
GI
~ 2000t------------------r---D~1
~ 1 5 0 0 t - - - - - - - - - - - - - - - - - + - - - - - T 1
·c
~ 1 0 0 0 t - - - - - - - - - - - - - - - - - t i r - - - - - 1
500t--------------~~~-----1
Penalty Value (C)
Fig.l0.S. Training time when using different optimizers on the hyperspectrai image
247
Thus, there appears to be no appreciable difference between the performances of the two optimization methods for the classification of both multi
and hyperspectral data. Either of these can be employed to perform SVM based
classification.