基于进化神经网络的RNA二级结构预测方法  

An evolutionary neural network approach to predict RNA secondary structure

作　　者：牟超[1] 何静媛[1] 石杨[1] 

机构地区：[1]重庆大学计算机学院,重庆400030

出　　处：《四川大学学报（自然科学版）》2014年第1期64-68,共5页Journal of Sichuan University(Natural Science Edition)

基　　金：中央高校基本科研业务费资助项目(CDJRC10180003)

摘　　要：本文提出了一种基于进化神经网络预测RNA二级结构的方法.该方法利用进化算法优化了神经网络结构和权值,改善了传统BP人工神经网络容易陷入局部最优等缺陷.实验结果表明该算法能够获得一个最优结构和权值的神经网络,利用该网络去预测不带伪结的RNA二级结构能获得理想的效果,敏感性和相关系数都有所提升,证明了该改进算法的可行性.A novel RNA secondary structure prediction approach based on optimized Neural Network （NN） is proposed. The structure and the weight of NN are optimized by evolutionary algorithm. Through optimizing the structure and weight, the deficiency such as trapping into the local area of tradi- tional NN is improved. Experiments show that this algorithm can achieve an optimal structure and weight of the NN. Using the NN to predict RNA secondary structure can obtain ideal effect. And the sensitivity and correlation coefficient of RNA secondary structure are improved. Simulation validates feasibility of the proposed approach.

关 键 词：RNA二级结构 预测 神经网络 进化算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]