[SciPy-User] annealing setting problem

scipy.optimize scipy.optimize@googlemail....
Wed Nov 24 11:20:33 CST 2010


Hi folks,

i have a problem with finding the right startsettings for anneal_fast.

My task is to optimize 6 variables.
2 of them should be optimised in a valuerange from about 10 to 40 (value1, value 2)
an the last 4 in a valuerange from about 0.00 to 0.05. (value3, value4, value5, value6)

Because I have 6 variables you see, that I need a global optimization.
The implementation is working fine. First I have tried the fmin Funktion but the range which was computed was way to small.
e.g.: value 3 was computed in an range from 0.01 (where it started) to 0.011
The same problem whith the other values. I know why. My startingpoint was to good and it found a local minimum. With a different startingpoint it found an other minimum.

So thats why I using annealing now. In special annealing_fast.
But Anneal_fast did not respect my given bounds (lower and upper). While I am searching for a value in range 0.00 to 0.05 it is computing in -10.00 to 30.00 even though the target value is way worse.

Then I was trying to find the right parametersetting. I vary T0 from 0.2 to 1.2 and dwell from 50 to 5000. After 20 different settings I could not see a correlation.

Thats why I hope one of you could help me to adjust my optimization. Does anyone already computed variables in my range with this method? Which parameter are the best (or even good) for my task?

Thankful kind regards,
Marius


More information about the SciPy-User mailing list