摘要
目的改进传统的BP算法,建立有效的药物活性识别模式。方法用遗传算法(GA)优化误差反向传播(BP)算法,两者结合构成混合算法。结果有效地解决了常规BP算法收敛速度慢、易陷入局部极小和GA算法独立训练神经网络速度慢等缺点。结论GA-BP结合构成的神经网络是一种有效的药物活性识别模式,可获得满意的预测结果。
Objective To improve traditional BP algorithm, to build effective the pattern recognition of the activity of medicament. Methods To optimize error backward propagation algorithm by using generic algorithm, to combine these two methods together to create hybrid algorithm. Results solved the problems faced by ordinary BP algorithm, i.e. slow convergence rate, easy to fall into local minima and slow independent training speed. Conclusion the hybrid GA-BP neural network is an effective recognition pattern of the activity of medicament and, satisfying forecasted results can be gained by using this method.
出处
《中国现代药物应用》
2008年第11期11-13,共3页
Chinese Journal of Modern Drug Application
关键词
BP神经网络
遗传算法
模式识别
BP neural network
Genetic algorithm
Pattern recognition
