Quadratic Program

A quadratic program is a quadratic optimization convex function (defined by the matrix \(Q\)) with a linear inequality constraint. The goal is to find the paramters \(\theta\) that minimizes the optimization function:

\[ f(\theta) = \theta^T Q \theta + c^T \theta + \alpha, \qquad A\theta \le b \]

• If \(Q\) is positive definite then there exists only one local extrema.

Simplex Algorithm

Sequential Minimal Optimization (SMO)

• Used by LIBSVM

Coordinate Descent

• Used by LIBLINEAR