基于模拟退火的Levenberg-Marquardt算法在神经网络中的应用

Application of ANN Based on Simulated Annealing Levenberg-Marquardt Algorithm

针对一般BP网络存在的一些缺陷,首次提出了利用基于模拟退火的Levenberg-Marquardt算法.在相同的初始条件下,用基于模拟退火的Levenberg-Marquardt算法的神经网络和Levenberg-Marquardt算法进行了比较,得出前者的特点和优点:收敛于全局最优解。一般函数逼近的实现表明,提出的算法是可行的,有效的。

Aimed at som limitation of ordinary BP neural network, an algorithm based on simulated annealing Levenberg-Marquardt algorithm applied to the neural network which predicts originally. On the same initial conditions, the neural network based on simulated anneling Levenberg-Marquardt algorithm with the neural network based on Levenberg-Marquardt algorithm are compared, and the characteristic and excellence of the former： higer precision and global optimization are explained. Experiments include normal functon show that the algorihm in this paper is practicble and effecive.