The Supporting information document contains detailed information regarding protocols and mathematical methods.

It is divided in the sections: Notes; Relation with Bayesian Model Selection; Image uniformation and de-uniformation; How many landmarks are too many?; Likelihood and evidence of the normal distribution; Likelihood and evidence of shape coordinates; Likelihood and evidence of texture coordinates; The concatenated code representation; Details of the classification algorithms; Results of the gender classification task; Different regularisation schemes; Visualisation of the eigenvectors of the concatenated code.

(PDF)