246
10: Pakorn Watanachaturaporn, Manoj K. Arora
Different Optimizers using Linear Kernel
200
180
160
~ 140
E 120
j:: 100
C>
t:
80
';::
'iii 60
...
I40
20
H -----CD --B- LSVM!
I
I
I
I
Y
~
~9f
I,",,",,",,",,",,",,",,",,",,",,",==~
L>...d
Penalty Value (C)
Fig. 10.5. Training time when using different optimizers on the multispectral image
0.70
0.60
~ 0.50
~ OAO
j::
~ 0.30
~
~ 0.20
0.10
I~
""
Different Optimizers using Linear Kernel
c:r
~
~
c:r
~~~X>
A
~~
/
""
V~""
L>.
f()
'V
V'
_____ CD --B- LSVM
Penalty Value (C)
Fig. 10.6. Testing time when using different optimizers on the multispectral image
and 10 3 (Fig. 10.7). For larger values of C, while the accuracy of the CD method
remains the same, it drops for LSVM.
In terms of training and testing times (Fig. 10.8 and Fig. 10.9), both the
optimizers take very little time and are very close to each other up to a C value
of 5, where the accuracy reaches its maximum. For values of C greater than 5,
the LSVM takes longer time than the CD method. But this becomes immaterial
since the maximum value of accuracy has been achieved at C == 5.