摘要
主要分析了神经网络和遗传算法的特点和存在的一些缺陷,研究了遗传算法和BP神经网络学习算法相结合的相关技术,设计并实现了一个基于遗传算法的BP神经网络算法BP-GA,已应用于肺癌早期细胞病理诊断系统中.实验结果表明,该算法具有较强的收敛性和鲁棒性,其应用效果很好.
The model of neural networks maily includes the following three elements: (1) the structure of neural networks including the layers and their relation; (2) the function of neural cell; (3) the algorithms of learning. There are many limitations implementing the neural network according to the integration of BP neural network learning algorithms with genetic algorithms. BP- GA, a BP neural network algorithm based on genetic algorithms is designed and implemented. This algorithm looks on every matrix of BP neural network as an unitary chromosome that is operated by the three major arithmetic operators including selection, crossover and mutation as the normal genetic algorithms. This mixed algorithm has been used in the diagnoses of lung cancer. The experimental results show that the algorithm has better convergence, robustness, and better effect in application.
出处
《南京大学学报(自然科学版)》
CAS
CSCD
北大核心
2003年第5期459-466,共8页
Journal of Nanjing University(Natural Science)
基金
江苏省自然科学基金(BK2001202
BK2002081)