Table 1.

Summary of classifiers compared.

Legend Figure 2	PCA components	Nearest-Neighbor (NN) Nearest-Mean (NM)	Distance metric	Distance metric D(y, x, Σ) =
NNC	–	NN	Correlation distance	$1-\frac{{\overrightarrow{y}}^T\overrightarrow{x}}{\left|\right|\overrightarrow{y}\left|\right|\left|\right|\overrightarrow{x}\left|\right|}$
NNE	–	NN	Euclidean distance	$\left|\right|\overrightarrow{y}-\overrightarrow{x}\left|\right|$
NMC	–	NM	Correlation distance
NME	–	NM	Euclidean distance	$\sqrt{\sum _n{(y_n-{\mu }_n)}^2}$
GNB	–	NM	norm. Euclidean distance	$\sqrt{\sum _n\frac{{(y_n-{\mu }_n)}^2}{{\Sigma }_{nn}}}$
NMC 8, 16, 32,…	8, 16, 32,…	NM	Correlation distance
NME 8, 16, 32,…	8, 16, 32,…	NM	Euclidean distance
GNB 8, 16, 32,…	8, 16, 32,…	NM	norm. Euclidean distance
LDA 8, 16, 32,…	8, 16, 32,…	NM	Mahalanobis distance	$\sqrt{{(\overrightarrow{y}-\overrightarrow{\mu })}^T{\Sigma }^{-1}(\overrightarrow{y}-\overrightarrow{\mu })}$

y, x: sample vectors of test and training data; y_n, x_n: components of y, x, μ: class mean (centroid) vector from training data; Σ_nn: diagonal elements of the covariance matrix Σ of the (pooled) training-data residuals (x−μ).