优化的径向基-循环子空间网络为药物定量构效关系建模

径向基．循环子空间回归(RBF-CSR)网络，保留了径向基-偏最小二乘(RBF—PLS)网络的优点，且可在更广的范围内选择最优模型，但仍存在着参数难以确定，计算量大等问题。对此，本研究兼顾网络模型的拟合与预测性能，采用具有高效全局搜优能力的优进遗传算法(EGA)优化网络参数，构建为EGA-RBF-CSR方法，并将其成功应用于苯乙酰胺类除草剂的构效关系(QSAR)建模，效果良好，显示出很强的学习能力，所建模型具有良好的预报性能和稳定性，并优于其他方法。

Optimized Radial Basis Functions-Cyclic Subspace Regression and Its Application to Quantitative Structure-activity Relationships

A new approach, radial basis functions-cyclic subspace regression (RBF-CSR), was proposed based on the analyzing radial basis functions-patial least squares (RBF-PLS). The approach has the merit of RBF-PLS, and it can select the optimal model in wider range. But selecting the parameter of the model is ~hard work. In this article, in order to solve the problem, an RBF-CSR base on eugenic evolution strategy ~genetic algorithm(EGA-RBF-CSR) was proposed, in which EGA was applied to optimize the parameter of RBF-CSR model for improving the performance of fitting and predicting. Finally, EGA-RBF-CSR was applied ~successfully to modeling quantitative structure-activity relationships. Compared to the other nonlinear models and linear models, The EGA-RBF-CSR model not only holds on fine learning ability but also gives better ~prediction performance and steady capability.