Gerhard Fuernkranz wrote: > [...] though I don't fully trust the A2B1 table either - the self-fit error > is IMO unreasonably low (I think I need to run a cross validation with > separate training and test sets to assess). > The residuals are indeed surprisingly low (max < 0.5, avg < 0.05 dE94), but a few brief CV runs (though just with a small data set) seem to confirm that the CV errors have a similar magnitude as for the corresponding CIELAB profiles. So likely this is still OK. Regard, Gerhard