Table 2.

JS divergences of the distributions of the CPC scores. For each model, we report the JS divergence between mixed set and test set (column $J{S}_m$) and the JS divergence between synthetic set and test set (column $J{S}_s$). The last four ${\mathrm{\Delta }}_{x,Z}$-like columns represent the improvement of MoGAN compared to the Gravity model on the mixed and the synthetic sets (columns ${\mathrm{\Delta }}_{m,G}$ and ${\mathrm{\Delta }}_{s,G}$) and the improvement of MoGAN compared to the Radiation model on the mixed and synthetic sets (columns ${\mathrm{\Delta }}_{m,R}$ and ${\mathrm{\Delta }}_{s,R}$)

Data	MoGAN		Gravity		Radiation		Rel. Improvement
	$J{S}_m$	$J{S}_s$	$J{S}_m$	$J{S}_s$	$J{S}_m$	$J{S}_s$	${\mathrm{\Delta }}_{m,G}$	${\mathrm{\Delta }}_{s,G}$	${\mathrm{\Delta }}_{m,R}$	${\mathrm{\Delta }}_{s,R}$
${\mathrm{NYC}}_{\mathrm{bike}}$	0.06	0.08	0.46	0.15	0.72	0.12	86%	49%	91%	37%
${\mathrm{NYC}}_{\mathrm{taxi}}$	0.09	0.11	0.53	0.14	0.83	0.15	83%	22%	89%	29%
${\mathrm{CHI}}_{\mathrm{bike}}$	0.14	0.16	0.29	0.25	0.56	0.26	51%	35%	75%	38%
${\mathrm{CHI}}_{\mathrm{taxi}}$	0.08	0.09	0.39	0.11	0.79	0.13	80%	21%	90%	30%