Methods Of Exponential Function Program Invariants Mining Based On GEP

Program Invariants is the stable character in software, which is used to describe the laws governing the operation of the program. The recent research work in Program Invariants using heuristics, divide into two stages-first preset invariant forms and then determine the invariant representation. Usually,Only the simple and limited number of invariant forms can be examined, so some important program features may be omitted. In addition, the detection process try to traverse all of the invariant collection, which is blind, time-consuming with low efficiency. But thanks to the GEP capability of function mining, the defect can be maked up by GEP. In fact, it has proved that GEP can resolve the problem of detecting standard form of polynomial function.This thesis continues the above idea and work on the research of invariant detection method of exponential function. First, function forms have to be detected through function-related collection by GEP; then calculate invariant representations; finally, detecting the invariant on data set. Usually Different program invariant forms have different characteristics. Beginning with the research of characteristics of GEP function form, the thesis find the main factors-a function symbol set and fitness function that effect the mining function, and at the end, it analyses the invariants of exponential function form. It was found that GEP can detect program invariants of lineal and exponential function efficiently, but to program invariants of non-lineal and exponential function, they can be detected by special ways. In a word, GEP has a better capability of form detection, and it can resolve the problem of program invariant of exponential function.The thesis summarizes and follows the former work, mainly works on the method to detect program invariants of exponential function form. It expands the capability of the current invariant technology on function detection and enhances the possibility of discovering more program invariants from program running data.