摘要
针对传统遗传算法 (GA)和人工神经网络 BP算法各自存在的不足 ,引入自适应机制的浮点数编码的遗传算法 ,并将其与 BP网中的梯度下降法相结合 ,进行混合交互运算 ,形成 GA- BP混合算法。该算法使网络具有较快的收敛速度和较高的逼近精度 ,能较好地解决综合多种地震信息进行薄互储层参数预测的精度和收敛速度问题 ,并通过实例验证了此方法的正确性和实用性。
Ageneticalgorithmwithadaptiveandfloating pointcodeis proposed to overcome disadvantages of the genetic algorithm and BP algorithm. This algorithm is combined with BP to give GA BP mixed algorithm which has higher accuracy and faster convergence speed. The new algorithm also provides improved predict accuracy of thin interbeded reservoir parameters. An example shows the validity and feasibility of this algorithm.
引文
[1]焦李成.神经网络系统理论[M].西安:西安电子科技大学出版社,1995.
[2]GrefenstetleJ.Optimization of control param eters forgenetic algorithms[J].IEEE Trans onSyst,Man andCybern,1986,16(1):122-128.
[3]黄秀轩,朱学峰.改进的自适应遗传算法[J].中国学术期刊文摘,1998,14(11):1415-1417.(HuangXiu- xuan,ZhuXue- feng.Modified adaptivegenetic algorithm[J].AcademicPeriodicalAbstracts ofChina,1998,14(11):1415-1417.)
[4]王建成,高大启,王静,等.改进的遗传和BP杂交算法及神经网络经济预警系统设计[J].系统工程理论与实践,1998,18(4):136-141.(WangJian- cheng,GaoDa- qi,WangJing,et al.Designing ofANN economic early warning system basedon improved genetic andBP algorithms[J].SystemsEngineering—Theory&Practice,1998,18(4):136-141.)