Fig. 13

Relative mean squared error vs. training size: a SVR2, \(\epsilon =10^{-4}\), b SVR3, \(\epsilon =10^{-4}\), c SVR rbf, \(\epsilon =10^{-4}\), d random forest, e boosting, f kNN, g VKOGA. Note that the validation set is fixed, \(N_\text {validation}= 500\). Blue curves represent training errors and red curves represent validation errors