摘要
对人工神经网络用于化学品生物活性预测中遇到的技术问题进行了系统的研究，提出了防止网络训练过拟合及局部最小的方法，对影响网络收敛的各技术参数的选择进行了探讨．确定的人工神经网络计算软件用于实际有机化学品生物活性预测，效果优于文献报道结果．
The artificial neural network (ANN) system based on back propagation algorithm has been studied for the prediction of biological toxicity of organic chemicals. The approach to prevent the overfitting and the local minimum, and the effect of ANN parameters on the network convergence have been discussed. The software of the artificial neural network system is applied to predict the bio activities of organic chemicals,the results are superior to ones published previously.
出处
《哈尔滨工业大学学报》
EI
CAS
CSCD
北大核心
1999年第1期101-103,106,共4页
Journal of Harbin Institute of Technology
关键词
神经网络
生物活性预测
QSAR
有机化学品
Artificial neural network(ANN)
Quantitative structure activity relationship(QSAR)
Bio activity prediction of organic chemicals
