Graphics Reference
In-Depth Information
(a)
(b)
Banana Original (0.8751,0.7476)
CNN (0.7729,0.8664,0.7304)
(c)
(d)
FCNN (0.8010,0.8655,0.7284)
IB3 (0.8711,0.8442,0.6854)
(e)
(f)
DROP3 (0.9151,0.8696,0.7356)
ICF (0.8635,0.8081,0.6088)
Fig. 8.4
Data subsets in banana data set (1).
a
Banana original (0
.
8751
,
0
.
7476).
b
CNN
(0
.
7729
,
0
.
8664
,
0
.
7304).
c
FCNN (0
.
8010
,
0
.
8655
,
0
.
7284).
d
IB3 (0
.
8711
,
0
.
8442
,
0
.
6854).
e
DROP3 (0
.
,
.
,
.
.
,
.
,
.
9151
0
8696
0
7356).
f
ICF (0
8635
0
8081
0
6088)
contains 5,300 examples in the complete set (Figs.
8.4
and
8.5
). It is an artificial data
set of 2 classes composed of three well-defined clusters of instances of the class
−
1
and two clusters of the class 1. Although the borders are clear among the clusters
there is a high overlap between both classes. The complete data set is illustrated in
Fig.
8.4
a.
