基于自适应遗传算法的神经网络字符识别

Character recognition based on adaptive genetic neural network

为了增强传统自适应遗传算法在进化初期的鲁棒性及避免早熟收敛,给出了一种改进算法,并利用这一算法来调整BP模型的网络权值与阈值,最后将新算法应用于字符的识别.仿真结果表明,新算法比BP算法具有更强的识别能力.

An improved algorithm has been given in order to make the robust characteristic of traditional adaptive genetic algorithm stronger during the initial evolutionary process and overcome premature convergence, then the neural network weights and threshold of BP model is adjusted by the algorithm. Finally, the new algorithm is applied to character recognition. Simulation results prove that the new algorithm has a stronger discernment than the BP algorithm.