摘要
为获取能快速分类的智能算法,在分析BP算法原理与缺陷的基础上,首先对标准的BP算法进行了改进,然后利用已改进的BPX算法优化遗传算法选优过程,提出了GA&BP混合算法,该算法兼顾了GA算法的全局收敛特性和BP算法快速的局部收敛能力,使算法既有较快的收敛速度又不易陷入局部解。仿真结果表明GA&BP混合算法的收敛速度、误差精度等主要性能指标有明显改善。
An improved standard Back Propagation （BP） algorithm is provided based on in-depth analysis of principle and defects of BP algorithm. Then ,the optimization process of Genetic Algorithm（GA） is improved. A novel hybrid GA＆BP algorithm is presented with the combinition of the GA algorithm and the improved BPX algorithm. The hybrid GA＆BP algorithm integrationg the merits of BP algorithm and GA has faster convergence speed and higher accuracy, and it can effectively avoid the local optimum phenomenon. Simulation resuhs verify the theoretical analysis.
出处
《计算机应用与软件》
CSCD
北大核心
2008年第4期60-63,共4页
Computer Applications and Software
基金
国家科技部国际合作项目(050296)
关键词
神经网络
BP算法
遗传算法
分类算法
Neural network BP algorithms Genetic algorithms Classifier algorithm
