Cascade Support Vector Machines with Dimensionality Reduction

Figure 1

Illustration of C-SVM and ECM learning scheme. In each level, the training set is divided into subsets. This procedure is repeated consecutively until the target size of training set for support vectors is reached. In case of ECM, the SVM learning process in the low-dimensional space returns low-dimensional support vectors. In each next level, the high-dimensional pendants are employed.