摘要
提出了一种基于前馈神经网络结构的适合于非线性预测的在线学习方法.这种方法吸收了最小二乘法和传统在线Bp算法的优点,具有收敛速度快、跟踪性能好、适用于非线性预测等特点.应用这种方法.对一些复杂的信号进行了一步预测,虫口模型、Heron方程和脑电信号的模拟结果验证了新方法的良好特性.
In this paper a real-time non-linear predictive algorithm based on feed-forward neural network was proposed. This method combines the advantages of both least square and Bp algorithm, so it has fast convergent speed and perfect trace performance.Specially, it is suitable for complex non-linear prediction. Some chaotic time series, such as Logistic mapping, Henon equation and EEG signal, were predicted by the proposed algorithm. The simulation results showed the good properties of the method.
出处
《复旦学报(自然科学版)》
CAS
CSCD
北大核心
1995年第3期262-268,共7页
Journal of Fudan University:Natural Science
基金
国家攀登计划认知科学(神经网络)重大关键项目
上海市科委重点基金
关键词
时间序列
预测
ARMA模型
浑沌
神经网络
on-line algorithm
ARMA(autoregressive moving average)model
chaos