Skip to main content

Table 3 Grand average training time \(\overline{T_t}\) and classification time \(\overline{T_c}\) of all 400 generated phantom maps using a dense array (100 samples) and two coarse (stimulation) arrays (\(3 \times 5\) and \(4 \times 6\) actuators, corresponding to simulation pooling sizes of \(15 \times 9\) and \(7 \times 7\))

From: Automatic hand phantom map generation and detection using decomposition support vector machines

Method

\(\overline{T_t}\) (ms)

\(\overline{T_c}\) (s)

\(\overline{T_t}\) (ms)

\(\overline{T_c}\) (s)

Dense array

 

OVA

OVO

 RS

35.0

15.9

54.9

33.7

 SS

28.6

15.3

47.9

32.7

 MP (2 \(\times \) 2)

84.2

17.5

79.3

39.6

 AL

300

16.1

278

38

 

DAG

BT

 RS

54.9

15.8

24.8

5.57

 SS

47.9

15.5

19.8

5.30

 MP (2 \(\times \) 2)

79.3

16.5

48.1

6.05

 AL

278

17

239

5.83

Hybrid coarse (stimulation) array, \(3 \times 5\) actuators

 

OVA

OVO

 MP \(15 \times 9\)

952

34.7

356

46.7

 

DAG

BT

 

356

35.8

301.9

9.13

Mechanotactile coarse (stimulation) array, \(4 \times 6\) actuators

 

OVA

OVO

 MP \(7 \times 7\)

348.3

25.2

196

45.0

 

DAG

BT

 

196

43.5

153

9.79

  1. RS random sampling, SS systematic sampling, MP majority pooling sampling, AL active learning, OVA one-vs-all, OVO one-vs-one, DAG direct acyclic graph, BT binary tree