These days, the boss asks me to have a look at the GP programming. Though busy now, I try to write something about the plans, the structures, the program flow and UML,etc. In my mind, a good plan is worth codes of several days.
The problem discription: Function set F={+,-,*,/,sin,cos}
Terminal set T=(x,y,z...)where the numbers are not set but can get from other place eg. database tables(such as column names)
The parameters set below:
Population size of 5000 individuals, number of generation is 1000, ramped half-and-half initialization, tournament selection of size 10, crossover rate equal to 95%, mutation rate equal to 0.1%,maximum tree depth for the initialization phase equal to 4, maximum depth for the crossover and mutation phases equal to 50.
Goal of function:
Automatically find the model from existed data sets and give the fitness curve.
So, my organization is divided into several steps:
First: understand fully the process of GP(especially the characters that will be used,eg.Adjucted Fitness, Encoding...)
Second:Java coding(effective and reusable with better UML figrues)
Third: later will be added...Let me have a rest, It is a long time since I last saw the web news....
Have a good time,every one.