提出了一种次梯度投影算法,解决凸可行问题,该算法在迭代过程中采用Armijo线搜索规则计算预测步长,且进一步给出一个校正步长规则,从而提高了算法的收敛性和收敛效果.最后给出了数值实例,表明算法的有效性.
Agradient projection algorithm for solving the convex feasibility problem(CFP) was presented and its convergence was shown under certain conditions.The algorithm uses the Armijo line search rule to calculate predictor step size and a step correction rule was given in the iterative process,which makes an accelerated convergence to the solution of CFP.Anumerical test was listed and the results generated are really impressive,which indicate the method is promising.