Fig. 2. (A) Example of a small 2D database with 12 points split among four families, indicated by their color. (B) Representation of the LOO-extrapolation to predict the values of ‘Family 1’, where the algorithm is trained with a LOO cross-validation. (C) Representation of our proposed LOGO-extrapolation to predict the values of ‘Family 1’, where the algorithm is trained to extrapolate to other families and is finally tested to extrapolate the values of that new family.