Notes. We calculated the errors by dividing the absolute difference between the MAP estimate and the literature value either by literature values (i.e. errors in percent units) or by the 1σ uncertainty of the literature value (i.e. errors in 1σ units). In the case of the effective temperature, the lσ uncertainty corresponds to the temperature interval of one subclass. For each network and parameter, we only used template stars whose literature values are within the training range of the network to calculate the errors.