Figure 5
From: Fast subcellular localization by cascaded fusion of signal-based and homology-based methods
Multi-class classification by perturbational discriminant analysis (PDA). (a) 2-dim data {x 1,…,x N } of 3 different classes in the input space, where N = 300. (b) N x N RBF kernel matrix K, where k i , j = exp{-|| x i - x j ||2/2}; each column of K is an N-dim vector k(x i ) in the empirical space K. (c) Projected data g(x i ) = ÂT k(x i ) + b on the PDA space where data can be easily classified by 1-vs-rest SVMs. (d) Decision boundaries produced by PDAproj+SVM.