论文部分内容阅读
为提高二级结构预测精度,试用神经网络集成法预测。针对BRNN网络结构复杂、收敛时间长、参数多的缺点,本文提出一种改进的新BRNN网络,删除BRNN左、右子网络的隐层,直接将输入连接到状态层。并采用BP改进算法中的弹性算法训练。以90条蛋白质序列共15 377个氨基酸交叉验证,仿真结果表明新网络可以有效地缩短收敛时间,新BRNN集成预测二级结构效果较好。
In order to improve the prediction accuracy of secondary structure, the neural network ensemble method is used to predict. Aiming at the shortcomings of complex BRNN network structure, long convergence time and many parameters, an improved BRNN network is proposed in this paper. The hidden layer of BRNN left and right subnets is deleted and the input is connected to the state layer directly. And use BP algorithm to improve the flexibility of training algorithm. The cross validation of 15 377 amino acids with 90 protein sequences showed that the new network can effectively shorten the convergence time and the new BRNN integration predicted the secondary structure better.