摘要
该文比较了神经网络与遗传算法的特点,提出了一种融合遗传算法和BP算法的神经网络算法设计。该方法采用了基于实数编码的改进遗传算法来替代随机设定神经网络的初始权阈值,然后由改进的LMBP算法在已由遗传算法确定了的搜索空间中对网络进行精确训练。仿真结果表明神经网络的逼近能力和泛化能力得到了综合提高,能够有效抑制遗传算法初期收敛的发生,确保了快速达到全局收敛,克服了传统BP算法精度低、收敛速度慢、容易陷入局部极小的缺陷。
This paper describes the characteristics of neural networks and genetic algorithm, presents a method of mixed neural network and genetic algorithm. The method adopts an improved genetic algorithm based on real - coded instead of the weight beginning with random value, then accurately trains the neural network with Levenherg - Marquadt algorithm. The simulation results indicate that the approximation capability and generalization ability of the network have been enhanced. Moreover, the premature convergence in genetic algorithm is restrained effectively and a rapid global convergence is guaranteed. The method also overcomes the shortcomings of traditional error back propagation algorithm for updating the weights of forward neural networks, such as the low precision of the solutions, the slow search speed and easy convergence to the local minimum points.
出处
《计算机仿真》
CSCD
2006年第1期161-164,共4页
Computer Simulation
关键词
遗传算法
神经网络
实数编码
算法
Genetic algorithm
Neural network
Real coding
Algorithm