I am modeling a data set using Eureqa. When I use this as the target string:
Target=f(cos(f1()-0.2618*MLT),cos(f2()-0.5236*MLT),Lpp,F10p7,F10p73,KpAv1,KpAv2,KpAv3,KpAv4,KpAv5,KpAv6,KpAv7,AeAv1,AeAv2,AeAv3,AeAv4,AeAv5,AeAv6,AeAv7)
this is the best solution that I get:
Target= 4.92406972 + 0.00076827*F10p7 + 0.00017782*AeAv4 + 19.75357498/(3.19461783 + Lpp) - 0.00015806*AeAv1 - 0.03793398*Lpp - 0.06707908*KpAv6 - 14.85211858*gauss(tanh(Lpp)) - 0.08737280*cos(0.51708340 - 0.26180000*MLT)
But when I used this as the target string:
Target=f(cos(f1()-0.2618*MLT),cos(f2()-0.5236*MLT),Lpp,F10p7,F10p73,KpAv6,AeAv1,AeAv4,DstAv1,DstAv2,DstAv3,DstAv4,DstAv5,DstAv6,PdynAv1,PdynAv2,PdynAv3,PdynAv4,PdynAv5,PdynAv6)
the solutions that I get are simpler and have greater error, no matter how many generations I use for the Eureqa run. This doesn't make sense, since the second target string includes all the terms in the original best solution. Since there is a greater space of possible solutions, a better solution should be found given enough computer time thrown at it.
Secondly, I tried to enter the solution as a prior solution, but I don't know how to do that, even though I read the example in the DataRobot documentation, which doesn't make sense to me. How would I write the prior solution above as an input to DataRobot?
The example in the Eureqa documentation is:
If Target = f0(x) * f1(x), and the prior model is Target = (x-1) * sin(2 * x), you will enter the prior solution as:
Target = (x - 1), f1 = sin(2 * x)
In that example it seems that f0 is specified as Target. Maybe that would work in that case, but in my case, you wouldn't know how f1 and f2 are related to f0.
Thanks in advance for any help.
日本語
