基于改进的自适应遗传算法优化BP神经网络

BP neural network optimized by improved adaptive genetic algorithm

针对Srinivas提出的自适应遗传算法种群前期进化较慢的问题,改进了自适应交叉率和变异率的计算方法,考虑交叉率和变异率与种群进化所处阶段的匹配,提出一种改进的自适应遗传算法;并将其应用于BP神经网络计算模型的优化,运用到汽车加油量计算中,通过比较标准BP网络、Srinivas提出的自适应遗传算法优化的BP神经网络和改进的自适应遗传算法优化的BP神经网络3种模型的计算误差,验证得出改进的自适应遗传算法优化BP神经网络的算法优于另外两种。

In the early stage of adaptive genetic algorithm （AGA） proposed by Srinivas, the speed of evolution is slow, which leads to reduction of the performance of algorithm. To solve this problem, based on improving the computing method of the probabilities of crossover and mutation and taking current stage of evolution into consideration, an improved adaptive genetic algorithm （IAGA） is presented. The new algorithm is applied to optimizing the calculation model of BP neural network, used to count the oiling quantity of vehicle. IAGA-BP is proved better than AGA-BP and standard BP by comparing the calculation error of these models.