提出了BP神经网络权值、阈值调整的双动量算法和学习率调整的批处理半恢复自适应调整法 ,基于此法计算了 15种含氢原子共价键的键长 ,计算键长的平均相对误差为 0 11%。结果表明 ,此二法的结合使用可改善网络的总体收敛性 Proposed the BP neural network weights, threshold-adjusted double-momentum algorithm and the learning rate adjustment batch semi-recovery adaptive adjustment method, based on which 15 kinds of hydrogen atom-containing covalent bond length calculated average bond length The relative error is 0 11%. The results show that the combination of these two methods can improve the overall convergence of the network