Skip to main content

Table 1 Evaluation of Algorithm 1 on the MNIST Dataset

From: Lγ-PageRank for semi-supervised learning

Digit

1

2

3

4

5

6

7

8

9

γ

7.0

3.0

7.0

3.2

3.2

7.0

7.0

3.2

4.2

\(\hat {\gamma }\)

5.45 (0.15)

3.10 (0.14)

6.41 (0.11)

4.92 (0.16)

3.20 (0.14)

6.04 (0.15)

4.98 (0.17)

4.40 (0.18)

5.08 (0.15)

\(h_{S_{gt}}^{(\gamma ^{*})}\)

0.065

0.166

0.035

0.141

0.131

0.011

0.052

0.116

0.135

\(h_{S_{gt}}^{(\hat {\gamma })}\)

0.073 (9e-4)

0.174 (8e-4)

0.041 (1e-3)

0.185 (4e-3)

0.148 (2e-3)

0.017 (1e-3)

0.074 (2e-3)

0.142 (2e-3)

0.149 (9e-4)

\(h_{S_{gt}}^{(1)}\)

0.175

0.248

0.216

0.258

0.233

0.107

0.203

0.215

0.285

  1. Mean values (95% confidence interval) are shown. The graph construction guidelines are provided in “Real world datasets” section. For the experiment, 500 realizations of labeled points and a grid of γ ranging from 1 to 7 with a resolution of 0.2 were used