一种级联过程神经元网络及其应用研究  

RESEARCH AND APPLICATIONS OF A CASCADE PROCESS NEURAL NETWORKS

作　　者：许少华[1] 何新贵[2] 

机构地区：[1]大庆石油学院计算机科学与工程系,大庆163318 [2]北京大学信息科学技术学院,北京100871

出　　处：《模式识别与人工智能》2004年第2期207-211,共5页Pattern Recognition and Artificial Intelligence

基　　金：国家自然科学基金(No.60373102)

摘　　要：针对连续系统输入过程可分为若干时间阶段,输出为一个即依赖于当前阶段系统过程输入,又与前一阶段系统状态有关的离散过程,提出了一种级联过程神经元网络模型.将连续输入信号分阶段处理,不同阶段系统输入输出映射关系用不同过程子网络描述.考虑过程神经元网络计算的复杂性,提出了一种基于函数正交基展开的学习方法,利用基函数的正交性,简化计算过程.文中给出了学习算法,并以油藏开发三次采油过程模拟为例验证了模型和算法的有效性.Aim at that the continuous system input procedure can be devided into some time stages and the system output is a discrete procedure that depends on system procedure input in every stage and relates to system output status of the previous stage, this paper gives a cascade process neural networks model. The continuous input signal will be decomposed in every stage, and relation of system input and output in different stage will be described with different sub-networks. In consideration of the computation complexity of process neural networks, the paper gives a learning method based on expansion of function orthogonal basis in which the computation procedure will be simplified using the orthogonality of basis function. The learning algorithm is given in this paper and the effectiveness of this model and algorithm is proved by tertiary oil recovery procedure simulation.

关 键 词：过程神经元网络 离散过程逼近 级联模型 函数正交基 学习算法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]