We create a quadratic approximation to the Langrangian function, linear approximation to the constraints, then we solve a quadratic problem in order to find the search direction. This method is able to perform a very fast optimisation. Obviously the first and second derivatives of the objective functions must exist.
