Skip to main content

Table 3 JS divergences of the distributions of the CPC scores with the Random Weighted model. For each model, we report the JS divergence between mixed set and test set (column \(JS_{m}\)) and the JS divergence between synthetic set and test set (column \(JS_{s}\)). The last two \(\Delta _{x,Z}\)-like columns represent the improvement of MoGAN compared to the Random Weighted model on the mixed and the synthetic sets (columns \(\Delta _{m,\mathrm{RW}}\) and \(\Delta _{s,\mathrm{RW}}\))

From: Generating mobility networks with generative adversarial networks

Data

MoGAN

Random Weighted

Rel. Improvement

\(JS_{m}\)

\(JS_{s}\)

\(JS_{m}\)

\(JS_{s}\)

\(\Delta _{m}\), RW

\(\Delta _{s}\), RW

\(\mathrm{NYC}_{\mathrm{bike}}\)

0.06

0.08

0.45

0.63

86%

88%

\(\mathrm{NYC}_{\mathrm{taxi}}\)

0.09

0.11

0.37

0.59

76%

82%

\(\mathrm{CHI}_{\mathrm{bike}}\)

0.14

0.16

0.4

0.56

64%

71%

\(\mathrm{CHI}_{\mathrm{taxi}}\)

0.08

0.09

0.37

0.55

79%

84%