Sequential Minimal Optimization (SMO) for SVR

SVR_SMO Create SVR model with SMO solver and different Kernels (linear, rbf, polynomial, sigmoid)

Run the example in file svr_test.m
SVR_SMO Create SVR model with SMO solver
vectors x(mx1), y(mx1) correspond to the set of points which y = f(x)
is a real function of the x real values. The SMO solver uses the
constants C, tol(tolerance), eps (epsilon), T (max.Iter). The choice of
the kernel is defined in type ('l' for linear, 'r for rbf, 'p' for
polynomial and 's' forsigmoid). Depending on the choice of kernel the
additionnalparameter will be used (gamma, offset and power).
The training result will be given as the alpha coefficients and the b
threshold.

Implementation from: Flake, Gary William, and Steve Lawrence. "Efficient
SVM regression training with SMO." Machine Learning 46, no.1-3 (2002): 271-290.