A support vector machine (SVM) is a multi-layer neural network where the first layer uses a large number of specialised non-linear functions in order to allow simpler leaerning in later levels. Often the non-linear functions form a family, called a kernal, for example they maybe foms of wavelet. Alongside other techniques such as reservoir ocmputing it is a way of creating non-linear diversity.
Used on pages 124, 144, 152, 154, 381
Also known as svm