论文部分内容阅读
采用自适应交叉变异、最优保存、局部寻优的遗传算法,避免了BP神经网络在训练过程中收敛于局部极小点的缺陷,并将其对神经网络的权值和阈值进行优化,从而提出了一种改进的混合遗传算法神经网络模型。该算法首先对一给定的神经网络结构,采用自适应交叉变异和最优保存策略对神经网络进行优化;然后采用局部寻优策略进一步克服神经网络学习算法的早熟问题。采用上述三种优化策略的神经网络模型对三元混合物溶液的物性和烟叶质量进行预测。试算结果表明,与实验值相比,预测结果良好。
The genetic algorithm with adaptive crossover mutation, optimal preservation and local optimization is adopted to avoid the defect that BP neural network converges to a local minimum during training and optimize its weight and threshold to the neural network An improved hybrid genetic algorithm neural network model is proposed. Firstly, the neural network is optimized for a given neural network structure by adaptive crossover mutation and optimal preserving strategy. Then the local optimization strategy is used to further overcome the premature problem of the neural network learning algorithm. The neural network model of the above three optimization strategies was used to predict the physical properties of the ternary mixture solution and the quality of the tobacco leaf. The trial results show that compared with the experimental value, the prediction result is good.