神经网络求解偏微分方程的分布式算法  

A Distributed Algorithm for Solving Steady-state Heat Transfer Problem with Neural Networks

作　　者：周欣[1] 刘伯安[1] 石秉学[1] 

机构地区：[1]清华大学微电子所,北京100084

出　　处：《微计算机信息》2006年第11X期224-226,125,共4页Control & Automation

基　　金：国家自然科学基金(69636030)

摘　　要：本文提出了一种人工神经网络(Artificial Neural Networks,即ANN)求解偏微分方程的分布式的算法。文章介绍了这种算法的必要性,以分布式的人工神经网络完成了对热传导方程的求解为例,验证了这种算法的可行性,并给出了不同条件下的实验结果。In this paper a new algorithm is presented which is described as a scalable distributed algorithm for solving steady-state heat transfer problem with artificial neural networks （ANN）. Their requirement and feasibility are mentioned. Partial differential equations（PDE） are solved by this distributed ANN to validate the feasibility, and laboratorial results in different conditions are given to make a comparison.

关 键 词：人工神经网络 分布式算法 静态热传导方程 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]