一种多轨道动态隧道技术训练BP网络算法

Algorithm of multi-trajectory dynamic tunneling training BP neural network

BP算法训练神经网络具有训练易陷入局部极小,收敛速度缓慢的缺点。将动态隧道技术运用到训练BP网络上,可以有效的改进BP网络易陷入局部极小的缺陷,但是传统的动态隧道技术训练BP网络算法在隧道方向具有不稳定性。提出一种多轨道动态隧道技术训练BP网络算法,在原基础上,增加了隧道搜索方向,考察搜索方向之间的相互影响,有效的改进了原算法的搜索效率。还对提出的新算法进行了性能分析,通过两种数据集进行了实验验证,证明其性能优于传统的动态隧道技术训练BP网络算法。

BP （Back Propagation） algorithm has some defects, such as converging slowly and immersing in local vibration frequently. The algorithm using dynamic tunneling technique to train BP neural networks is proved to have good performances in avoiding the local trap. But conventional dynamic tunneling training BP neural network algorithm has numerical instability, so a new dynamic tunneling algorithm with multi-trajectories is proposed in which the interaction between each trajectory of the tunneling system is introduced to improve search efficiency. The simulation results are provided for two different examples to demonstrate the performance of the proposed method in overcoming the problems of initialization, The performance of the conventional dynamic tunneling technique and the multitrajectory dynamic tunneling technique in training BP neural networks are given and compared.