aX = cov(x_,y_) aY = 0.5/n * ( ∑(x_^2) + ∑(y_^2) + sqrt( (∑(x_^2) + ∑(y_^2))^2 + 4 * cov(x_,y_)^2 ) ) What's meaning about (aX aY) Look like The PCA is not say the principal vector is (aX aY)