Fig. 5

The hyperplanes (dashed line) of SVM model separating patient groups via two voxels from right superior temporal pole (vertical axes) and left superior parietal (horizontal axes) with different kernel functions. (upper) The kernel function was in the form of Φ(x). Φ(x ^'), which produces a linear decision boundary having 117 out of 162 subjects were correctly classified. (lower) The separating line is non-linear since a polynomial kernel ((Φ(x). Φ(x ^'))⁴ was used to map the data. 120 out of 162 subjects were correctly classified