I am so proud of myself. I successfully applied PGEP to two different symbolic regression problems (the one in the original GEP paper and the one in the original PGEP paper). The solutions are so accurate I'm certain they are equal to the correct ones, as I'm getting minuscule numbers for the residual error.
This is especially awesome because I've had problems with duplicating other peoples results, but this time it seems like I'm able to find the correct functions. Especially when adding constants to the terminal set, which seems to have improve the quality of the solutions by several thousand times with only 1, 2, 3, 4, and 7.
Now if only I can replicate those results (or something similar) with BPGEP I will be able to actually start collecting data and all that.
Dude, thats awesome. I remember how duplicating results has been an issue. I have to say- Symbolic Regression sounds to me like a Deconstructionist Freudian/Lacanian school of literary criticism.
ReplyDelete