Top: the estimated photo-z versus the actual redshift, as determined by SPIDERz, for the AEGIS data set discussed in Sect. 6 with no morphological parameters included. The training set is formed from 700 galaxies and the evaluation set, for which the results are plotted, consists of the remaining 1912 galaxies. “Outliers” in a determination are defined by Eq. (1), shown as those outside of the two diagonal lines. Middle: photo-zs for the same galaxies, as estimated with the custom neural network as reported in Singal et al. (2011). Bottom: photo-zs for the same galaxies as estimated with the Lephare template fitting code (Ilbert et al. 2006). The template fitting method has a lower scatter for non-outliers, but a larger number of outliers than the SVM and neural network for these galaxies.