图形匹配算法在公式发现理论FDD中的应用

Application of graphics matching algorithm in formula discovery from data

传统的经验公式发现理论FDD虽采用启发式搜索,但随着函数库的扩大,仍存在函数的组合爆炸现象。根据复合函数曲线仍然具有原函数图形特征的性质,提出了从函数曲线的图形特征入手,采用图形匹配的新方法,将各类初等函数的曲线图形作为搜索库元素,用Hausdorff距离匹配法将目标函数曲线与搜索库元素进行匹配。实验结果表明,该方法能够快速有效地搜索到原型函数,提高了算法的搜索效率。

The traditional theory of the formula discovery from data used heuristic search, with the functions library expanded, there is still a combination of function explosion phenomena. Based on the composite function curve still has a primary function graphics features, the paper started from the function curve＇s characters, used the matching method of graphics, made the various primary function＇s curves as an element of the search library. Matching the objective functions and the curves in the library used the matching method of Hausdorff distance, experiments show that this method can search the prototype functions quickly and efficiently, the efficiency of the whole system algorithm is improved.