Skip to main content
. 2017 Feb 22;12(2):e0172578. doi: 10.1371/journal.pone.0172578

Table 4. 10-fold cross-validation of ambulatory SSVEP classification by the quantity of training data.

11,250 22,500 33,750 45,000 56,250 67,500
Low All Low All Low All Low All Low All Low All
CCA-KNN 79.51
(+3.04)
84.64
(+3.24)
79.28
(-3.11)
84.49
(-1.69)
79.16
(-6.43)
84.41
(-4.38)
79.29
(-8.56)
84.52
(-6.14)
79.36
(-12.09)
84.58
(-8.72)
79.46
(-12.87)
84.65
(-9.38)
NN 75.15
(-0.88)
77.81
(-3.59)
82.63
(+0.24)
84.61
(-1.57)
86.39
(-0.8)
87.93
(-0.86)
89.21
(+1.36)
90.35
(-0.31)
91.04
(-0.41)
91.96
(-1.34)
91.94
(-0.41)
92.80
(-1.23)
CNN-2 59.21
(-17.26)
68.31
(-13.09)
73.77
(-8.62)
79.78
(-3.4)
78.95
(-6.64)
83.99
(-4.8)
82.52
(-5.33)
85.77
(-4.89)
83.91
(-7.54)
87.75
(-5.55)
85.16
(-7.19)
88.36
(-5.67)
CNN-1 76.47 81.40 82.39 86.18 85.59 88.79 87.85 90.66 91.45 93.30 92.35 94.03

10-fold cross-validation of the ambulatory SSVEP classification when changing the number of training data with 7,500 test data. Low indicates subjects who have a low CCA accuracy (under 80%, i.e., subjects S3–7). In parentheses, the accuracy differences of the methods compared with CNN-1 are indicated.