I have some simple questions, which I hope someone can answer.
I am experimenting with the Galapagos evolutionary solver for my master's thesis. I've set up a model which is able to send and retrieve data to and from a structural analysis program (so that I have calculation data). One specific result from that calculation data is used for determining the fitness of my model.
- When using the two different solvers (the evolutionary solver and the simulated annealing solver) I seem to get different outcomes for the calculation. That is; the genome is exactly the same, but results differ. When manually setting the genes to the optimal genome, i get the same result as with the SA solver. Can anyone explain how the difference emerges? (Note: I actually already know what the optimal genome for my problem is for the test I'm doing right now)
- Depending on the slider setup with which I start, Galapagos tends to find different optima (either global or local). How can I best avoid finding a local optimum? Can this be done by entering a higher value for 'initial boost'?
- Which selection mechanism is used? Is it 'isotropic selection'?
- Can anyone refer me to good literature on how both genetic algorithms and simulated annealing works? I'm currently reading Genetic Algorithms by David Goldberg, but I'm not sure if it is all up to date.
I hope my questions are clear. Otherwise I'll be happy to upload a video explaining what I mean :)