Table 4.

Predicted log-likelihood of stationary experimental data under dynamical models. The rows contain measurements from a stationary distribution under the corresponding culture conditions (figure 4). The columns contain models obtained from independent dynamic experiments under culture conditions (figure 5). Each entry shows the log-likelihood of the model in the column applied to the data in the row. The highest likelihood for each dataset, the maximum in each row (highlighted in bold), should be achieved by a model trained on separate data under the same condition.

	LIF+BMP4	PD03	Chiron	2i
LIF + BMP4	−30 655	−39 091	−31 680	−45 5367
PD03	−59 980	−43 988	−62 506	−111 399
Chiron	−32 127	−48 926	−27 572	−35 164
2i	−30 494	−53 451	−24 947	−17 012