Table 3. Neural-Network Approximate Bayesian Computation
posterior parameter errors under the winning scenario AfrDE-EurDE, for
the ACB and ASW populations. For each target population separately, we
conducted cross-validation by considering in turn 1,000 separate NN-ABC
parameter inferences each using in turn one of the 1,000 closest
simulations to the observed ACB (or ASW) data as the target
pseudo-observed simulation. All posterior parameter estimations were
conducted using 100,000 simulations under scenario AfrDE-EurDE
(Figure 1 , Table 1 ), a 1% tolerance rate (1,000
simulations), 24 summary statistics, logit transformation of all
parameters, and four neurons in the hidden layer (see Materials
and Methods ). Median was considered as the point posterior parameter
estimation for all parameters. First column provides the average
absolute error; second column shows the mean-squared error; third column
shows the mean-squared error scaled by the parameter’s observed variance
(see Materials and Methods for error formulas).