摘要
根据变量间的实验数据来发现其内在联系是科学研究及工程设计中经常涉及到的问题。采用人工智能和曲线拟合相结合的经验公式发现系统,较传统的数据拟合技术更为直观,且有效地避免了传统数据拟合技术中系数行列式元素微小变化引起解的显著变化的"病态"问题。针对经验公式发现系统搜索方向判断标准比较单一的问题,提出了对其搜索方向判断标准的扩充与改进方案,引入误差方差作为判断搜索方向的标准,实践证明:搜索方向判断标准的改进能够使系统较为准确地找到正确的搜索方向,发现变量之间的关系,提高了系统的可用性。
It often needs to find the relation of variables according to the experiment data. The result by way of Formula Discovery from Data （FDD） which is based on Artificial Intelligence and the Curve Fitting is more intuitive than Traditional Data Fitting technique. And FDD can solve the problem in Traditional Data Fitting that it can bring large change in polynomial ＇s solution because of little change in polynomial＇s elements. The improvement method of search direction criterion is proposed for the problem that the search direction criterion of FDD is relatively simple. The Deviation variance is introduced as the criteria to judge the search direction. From the experiment, the Improvement of search direction criterion can accurately find the right search direction and discover the relationship between variables. It improves the usability of FDD.
出处
《微计算机信息》
2011年第4期238-239,218,共3页
Control & Automation
关键词
经验公式发现系统
人工智能
数据拟合
搜索方向
误差方差
Formula Discovery from Data
Artificial Intelligence
Data Fitting
Search Direction
Deviation Variance
