A projection subgradient algorithm for the Lagrangian dual problem of the relaxed quadratic problem is employed to general lower bounds of the optimal value for the original problem.

The algorithm relaxed the restricted conditions by fuzzing the element of partition matrix in the iterative process.