欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
目前常用的物体识别方法,其过程非常复杂,信息量和计算量都很大。结合遗传算法的神经网络方法,充分利用GA的全局搜索能力、BP算法的局部搜索能力和鲁棒性强的特性,提出了一种用遗传算法全局优化神经网络拓扑结构和网络权值的新编码方案进行物体识别方法。仿真结果表明,该方法既解决了BP神经网络对初始权值敏感和容易局部收敛的问题,又加快GA-BP网络的收敛速度,提高收敛精度且识别率较高,从而验证了该方法的有效性。
英文摘要：
Algorithms of recognition are complex, and the information and computation are large at present. Neural network based on genetic algorithm makes fully use of GAps global searching and BP networks local searching, and it has good robustness. The new coding scheme of the global optimization of the topology and weight distribution of neural network fusing with genetic algorithm is proposed. The simulation results demonstrate that the method not only overcomes the sensitivity to the initial weight and the local convergence, but also improves the convergent rate and convergent precision. Moreover, the recognition rate is very high. So the efficiency is proved.
同期刊论文项目
同项目期刊论文
期刊信息
位置：