【摘要】 蛋白质结构预测在生物信息学研究中占有重要地位,而蛋白质二级结构预测是蛋白质结构预测的关键步骤。针对标准BP算法存在的缺点,讨论采用几种不同的改进BP神经网络来实现蛋白质二级结构的预测,运用MATLAB语言实现各种改进算法的初始化和训练。并分析比较了它们对蛋白质二级结构预测精度的影响。实验表明,遗传算法结合动量法与学习率自适应调整策略的BP算法可获得较高的预测精度。更多还原

【Abstract】 Predicting protein structure plays an important role in bioinformatics. In this process the pivotal step is predicting from protein sequence to secondary structure. Considering the disadvantages of standard BP arithmetic, it introduces several improved BP neural networks to predict protein secondary structure, which apply MATLAB to initialize BP neural networks, and analyses the precisions of prediction. The experiment has proved that a high precision can be achieved by the improved BP network which combines genetic algorithm with method of momentum and strategy that is adopted for adjusting learning efficiency.更多还原