Cascade Support Vector Machines with Dimensionality Reduction

Figure 2

Comparison of SVM learning results of (a) a classic SVM trained on a set of patterns (XOR data set), (b) an SVM trained with the support vectors of an SVM that has been trained on the 1st half of the XOR data set, (c) an SVM trained on the support vectors, 2nd half of XOR data set, and (d) the C-SVM trained on the support vectors of both SVMs.

(a) SVM trained with all patterns

(b) SVM trained with support vectors of 1st half

(c) SVM trained with support vectors of 2nd half

(d) C-SVM trained with support vectors of two SVMs