which doesn't exist in the actual problem spec. If galapagos is allowed to change the column position in all possible directions, it is less likely to get stuck in some local optimum.
Let's assume that (all other things being equal) column 20 would yield the best possible answer. The current state of the system though is at column 26, which is pretty good too, just not as good. Galapagos is more likely to 'mutate' the state a little bit instead of a lot, so it'll explore the columns near 26. However 20 isn't near 26 at all, only 25 and 27 are nearby, and maybe 24 and 28. But they'll all worse answers, so after sampling in those directions GP will abandon that as fruitless.
If however you specify the columns using two variables, then the columns near 26 are 20, 25, 27 and 32. That's a far richer space to explore which much better approximates the real problem.…
Branch Address in the form {0;1;2;3;4;5;6;7;8;9;.......;N}
Try Grafting the input (I) Create Path by right clicking on it and selecting graft from the context menu…
en 3 of them, and one poolyline between two of them.
It would also be very nice if i could control it so that only the successive ones can be connected
so if {0:0:0} has 8 points and {0:0:1} has 8, as do {0:0:5} and {0:0:6} i would like to have this as two polylines, not one continoous that would in this case jump three branches (or curves that are shorter).
Does this make any sense?…
Added by Dusan Bosnjak at 2:08pm on September 28, 2009