【摘要】 针对误差反向传播(BP)算法训练速度慢和易于陷入局部最小值的缺点,利用BP算法监督学习特点,模拟退火算法(SA)在局部极小处的概率突变性,有效结合BP和SA算法,提出一种BPSA混合学习策略。将其应用于GPS高程拟合,并以实例验证了该算法的有效性。更多还原

【Abstract】 As the error back propagation(BP) algorithm provides a slow training and is easy to fall into local extremum,a hybrid strategy named BPSA which combines the surpervising learning property of BP with the probabilistic jumping property of simulated annealing(SA) algorithm is presented.It is applied to the GPS height fitting.A practical sample is set to justify its availability.更多还原