Table 5.

Discretization error of the synaptic-input statistics for 1- and 2-bin discretization.

x	〈μ〉	〈σ2〉	〈δμ2〉	〈δ(σ2)2〉
ε1(x)	0	j	$j·\frac{1+f}{k+f}$	$j·(2+j)·(1+2\frac{1+f}{k+f})$
ε2(x)	0	0	0	$\frac{2j^2}{4j+\frac{k+f}{1+f}{(1+j)}^2}$

Discretization error ε_Nbins(x) as defined in Equation (25) for the four ensemble measures Equations (21)–(24) (columns), for networks with N_bins = 1 and N_bins = 2 weight bins (rows). The parameters j: = 〈δw²〉/〈w〉², f: = 〈δν²〉/〈ν〉² and k: = 〈δK²〉/〈K〉 denote the squared variation coefficients of the synaptic weights and the firing rates, and the Fano factor of the in-degrees, respectively.