极限学习机优化方法在蛋白质折叠类型识别中的应用

摘要

传统的机器学习方法在处理蛋白质折叠类型识别问题时需要花费大量的时间来调节最佳的参数.利用一种新的极限学习机(Extreme Learning Machine,ELM)分类优化方法(Extreme Learning Machine for Classification,ELMC)对蛋白质折叠进行识别,仅需调节很少的参数值就可达到很好的测试精度.与支持向量机(Support Vector Machine,SVM)和推荐相关向量机(Relevance Vector Machine,RVM)相比,ELMC能获得更好的泛化性能,而且在寻找最优解的训练时间比较上,ELMC比SVM平均要快35倍,比RVM要快12倍.%With traditional machine learning methods, one may spends a lot of time adjusting the optimal parameters in tackling the problem of protein fold recognition. A new optimization method of extreme learning machine for classification ( ELMC ) is used to recognize the protein fold, one can only adjusts few parameters to achieve good enough testing accuracy. Compared to support vector machine (SVM)and relevance vector machine (RVM) , better generalization performance can be obtained by extreme earning machine for classification. In the comparison of training time in finding the optimal solution, ELMC is 35 times faster than SVM averagely and is 12 times faster than RVM averagely.