摘要
提出了前馈网络目的规划算法。与通常ＢＰ算法相比，该方法进行了三个方面的改进：（１）准则函数的改进；（２）网络灵敏度的降低；（３）领域先验知识的运用。理论分析及大气中ＳＯ２浓度预测应用研究表明该方法有效地改善了前馈网络泛化性能，提高了预报精度。
One of the important characteristic of feedforward neural networks is their ability to generalize the input/output behavior of functions based on a set of training examples.Yet many aspects of the problem of the improving generalization in feedforward neural networks has not been studied well.The goal programming method for neural network in air pollution prediction is proposed in this paper.Three enhancements are made in this method,those are: (1) improvement of objective function; (2) uses of prior knowledge of air pollution; (3) decrease of the sensitivity of the neural network.Theoretial analysis and case study show that both generalization performance and the precision of forecasting of the neural networks are improved effectively.
出处
《系统工程理论与实践》
EI
CSCD
北大核心
1998年第2期56-61,共6页
Systems Engineering-Theory & Practice
基金
国家自然科学基金
武汉市晨光计划资助
关键词
前馈网络
目的规划
大气污染预测
二氧化硫
feedforward neural networks
generaliztiion performance
goal programming
air pollution prediction
