1: \begin{abstract}

2: In this paper, proximal two-stage algorithms for quadratic programming problem subject only to box constraints are presented. Program of this type, is denoted by BQP, which plays an important role in many optimization models. We apply proximal point method to solve non-strictly-convex BQP problem, which has proved to be globally convergent and have linear convergence rate at local. Moreover, we show that the proximal point method is essentially linear iteration if the active-set is fixed. According to this property, we give an estimate of the linear convergence factor and present an accelerated proximal point algorithm. For strictly convex BQP problem, the accelerated proximal gradient method is first implemented to get a approximate solution; secondly, we generate a parametric programming with the approximate solution and track the solution path of the parametric programming with a simplified parametric active-set method. Moreover, in the path tracking steps, we improve the parametric active-set method by introducing a sorting technique to update the cholesky decomposition. The numerical experiments on randomly generated data set and real-world data set show the accelerated proximal point algorithm can greatly reduce the number of the iterations and the two-stage algorithm is efficient.

3: 

4: \end{abstract}

5: