摘要
为了建立HIV蛋白酶抑制剂QSAR的优良模型,本文采用粒子群优化法搜索支持向量机的多参数复杂模型空间,以此形成最优支持向量机。通过与传统的梯度下降法、网格搜索法等模型选择方法的比较,采用并行计算的基于PSO算法的最优支持向量机法在模型精度及稳定性、搜索效率等方面都有优良的性能,实例测试也表明所建QSAR模型,有良好的泛化能力,所建模型对研究HIV药物有重要促进作用。
The optimal support vector machine was proposed in this paper by applying particle swarm optimization searching the complex multi-variable space in SVM model, in order to modeling the better QSAR. Compared with gradient descending, grid searching, the PSO- SVM algorithm using parallel computing shows better performance in model accuracy, model stability, and computing efficiency. The PS0-SVM algorithm was successfully employed in modeling the HIV-1 protease inhibitors which has more generalization ability.
出处
《计算机与应用化学》
CAS
CSCD
北大核心
2007年第11期1475-1478,共4页
Computers and Applied Chemistry