基于改进牛顿算法的蛋白质二级结构预测

Protein Secondary Structure Prediction Based on Improved Newton Algorithm

主要介绍构造性机器学习方法即改进牛顿算法在蛋白质二级结构预测中的应用。针对标准BP算法存在的缺点,讨论用迭代矩阵替换二级微商来改进牛顿算法,实现蛋白质二级结构预测。实验表明,采用基于概率的Profile编码方式,改进牛顿算法正确率可以高达73.68%,与其他预测方法相比有较好的准确性。

Protein secondary structure prediction based on improved Newton algorithm which is a method of structure machine learning is mainly introdued. Considering the disadvantages of standard BP algorithm, it discusses improved Newton algorithm to predict protein secondary structure by using matrix iteration to exchange the secondary differential quotient. The experiment results indicate that the degree of accuracy of prediction results achieves 73.68% by using Profile encoding and it is better than any others.