Fig. 2From: Rigorous optimisation of multilinear discriminant analysis with Tucker and PARAFAC structuresObjective function values. Objective function values for one, three, and five components. Scatter ratio objective function (5) values are shown as full lines while the matrix ratio objective (6) is shown as dashed lines for three random initialisations. Top: Stekelenburg&Vroomen data for the CV fold with subject 5 left out. Bottom: subject B from the BCI data. Note the log scale of the y-axis in the upper row and the linear scale in the bottom rowBack to article page