摘要
针对传统BP算法存在的收敛速度过慢、易陷入局部极小、缺乏统一的理论指导网络结构设计的缺点,分析了一般的改进算法在神经网络优化过程中存在的问题,从蚁群算法和BP算法融合的角度上,并引入了放大因子,提出一种综合改进的BP算法。该算法引入放大因子改善BP算法易陷入局部极小的情况,结合蚁群算法用于指导网络结构设计,并极大地改善了收敛速度过慢的问题。最后,将改进的BP算法与传统BP算法进行应用于煤矿瓦斯预测。通过对实验结果的分析,从时间和正确率上都表明改进的BP算法要优于传统的BP算法。
As for shoaonmings of standard BP algorithm such as slow convergence,easily trapped into local minima and no unified theory to guide how to design neural network,anelyzed the problems of original improved BP algorithm,from the perspective of combing ant coimly optimization with BP algorithm introduced a new enlarge factor,thus proposed a new impwved BP algorithm. The algorithm is in- troduced into the amplification factor to improve the BP algorithm is easy to fall into local minima,combined with the ant colony algo- rithm is used to guide the network stngture design,greatly improve the convergence speed. On these basis,the proposed BP algorithm and classical BP algorithm were applied to the prediction of coal mine gas concentration. By analysing the experiment, the results show that the improved BP algorithm indeed is more efficient than classical BP algorithm from time and accuracy.
出处
《计算机技术与发展》
2012年第8期163-166,共4页
Computer Technology and Development
基金
山西省自然科学基金项目(2008011028-1)
山西省科技攻关项目(20100322003)
关键词
BP算法
蚁群优化算法
神经网络
BP algorithm
ant colony optimization
neural network
