摘要
介绍了一种将组合神经网络用于聚合物质量预测的方法.由定量数据建立的单一神经网络模型往往缺乏泛化能力,而使用组合神经网络模型则可以显著改善模型的泛化能力.由于在建立组合神经网络模型过程中,合适的组合权重对模型是否具有良好预测性能是非常重要的,因此采用了岭回归方法来选择合适的组合权重.所提出的方法已成功应用于PVC颗粒特性的预测研究中。研究结果表明,与单一神经网络模型相比,组合神经网络模型具有更佳的模型预测精度和鲁棒性.
Inferential estimation of polymer quality using stacked neural networks (SNNs) is studied. A single neural network model developed from a limited amount of data usually lacks generalization capability. Model generalization capability can be significantly improved by using SNNs model. Proper determination of the stacking weights is essential for good SNNs model performance, so determination of appropriate weights for combining individual networks using ridge regression is proposed. The proposed technique has been successfully applied to the studies of prediction for properties of PVC grains. The results obtained demonstrate significant improvements in model accuracy and robustness, as a result of using SNNs model, compared to using single neural network model.
出处
《科技通报》
北大核心
2004年第4期298-302,共5页
Bulletin of Science and Technology
基金
国家杰出青年科学基金(20125617)
浙江省自然科学基金资助项目(200024)
高等学校重点实验室访问学者基金资助项目
关键词
高聚物工程
神经网络
预测
岭回归
组合泛化
PVC颗粒特性
polymer engineering
neural networks
prediction
ridge regression
stacked generalization
properties of PVC grains
