Figure 2

Parameter optimization for the gene expression networks. The standard deviation σ is given as a function of the total number of parameters p. The five curves correspond to the five model structures of eqs (5-9): dashed line: $m_{NC}^{pol}$; thick solid: $m_{NN}^{exp}$; dashed-dotted line: $m_{NC}^{exp}$; dotted line: $m_{CN}^{exp}$; thin solid line: m _lin . The number of parameters (p) depends on the connectivity per class (q) and the number of classes (C): p = (2q+4)C for $m_{NC}^{pol}$, p = (2q+5)C for $m_{NN}^{exp}$, p = (q+4)C for $m_{NC}^{exp}$ and $m_{CN}^{exp}$, and p = (q+1)C for m _lin . The value of the minimum connectivity q^m necessary to yield good data reproduction is indicated by a large circle. (a) Embryonic time series; the number of parameters p shown corresponds to connectivities q between 1 and 4; q^m = 3; (b) full time series; the number of parameters p shown corresponds to connectivities q between 1 and 12; q^m = 7.