Hi,
I’m using Grasshoper for optimization problems. There are numerous plug-ins that allow to reach optimized solutions through the use of genetic algorithms and I use one of these. I often struggle with the need to further optimize the solution of a previous simulation (because more epochs would have been needed). The ideal solution would be to be able to keep in memory the best solutions from the first simulation and use these as a starting point for a second simulation. So far I haven’t found any tools that allow me to use this approach. Do you have any advice to solve this problem?
Thanks and best regards
Alessia
Post an example file. It is a bit difficult for someone to guess which tools you are using and for what particular problem. Maybe there is a way to loop the solution, maybe there is a way to code it. But I doubt there is an out of the box general solution to what you want.
I can’t post the file I use for my simulations because it also uses external softare for carrying out calculations. However I posted a file in which the goal is to minimize the distance between the vertices of a tringle and in which the optimization tool is the same one that I use in my simulations. I try to give an example to better explain my problem: I launch a simulation A with 500 epochs and at the end I realize that 500 epochs are not enough for the level of optimization I want to achieve. I should then restart a new simulation B by setting a higher number of epochs: this involves the loss of all the results of simulation A and therefore I would have to waste hours of computation time. The ideal solution would consist in starting simulation B not from a random condition but from the solutions obtained with simulation A. In practice, simulation B would be placed after A without having to restart from the beginning.Example.gh (13.0 KB)
The trouble with Wallacei is that the first generation is always random. I don’t know how one can force it to take as generation 0 some specific values. I also haven’t seen any other evolutionary engine that permists you to do it. Maybe someone on the forum has more experience with this. But I am in doubt it can be done.
Galapagos provides these two options (that sound relevant, if I understood the problem):
1 Like
Oh… But then wouldn’t it repeat generations already disgarded in the previous calculation?
But starting from that initial set, does Galapagos create the first generation with random variations?
