基于遗传算法的BP网络及其非线性逼近能力研究

Research of GA-based BP Network and its Ability for Nonlinear Approach

提出了一种基于遗传算法(GA)的BP神经网络模型优化方案,指出了遗传算法和标准BP算法各自的优缺点。首先采用自适应交叉概率的遗传算法优化网络的权值,在进化结束时,能够寻到全局最优点附近的点。在遗传算法搜索结果的基础上,利用局部寻优能力较强的梯度下降法,从此点出发,进行局部搜索,进而达到网络的训练目标。仿真表明与单一的梯度下降法比较,混合优化算法的收敛速度快,逼近的效果好,因而所给出的算法可行有效。

With the merits of the genetic algorithm and the simple BP algorithm,a BP network model based on GA is presented. Adopting the adaptive cross probability,the GA is to optimize the weight. Next,at the end of evolution,a solution near the global optimum will be found. Then,the global optimum in the local area can be found by adopting gradient descent algorithm. Compared with the gradient descent algorithm,the simulation shows the mixed algorithm is feasible and effective.