197352.fig.003a
(a) Gaussian kernel and 2 features
197352.fig.003b
(b) Mexican hat kernel and 2 features
197352.fig.003c
(c) Gaussian kernel and 4 features
197352.fig.003d
(d) Mexican hat kernel and 4 features
197352.fig.003e
(e) Gaussian kernel and 20 features
197352.fig.003f
(f) Mexican hat kernel and 20 features
Figure 3: Sample standard deviations of average classification accuracy rates when FLD classifier is used only (cyan plots) and when a feature extraction method plus FLD classifier is used for 100 Monte Carlo simulations. The considered different types of feature extraction methods are PCA (blue plots), STIWK PCA (green plots), MTIWK PCA (red plots), and MDWK PCA (black plots) for the data simulated with different values of 𝜎 𝑥 .