QN are a family of methods for optimization, based on Newton methods. They, however, do not require the computation of the Hessian matrix.

Since they do not need to compute Hessian, these methods are computationally cheaper, as well as require less memory footprint.