Questions in GALAPAGOS !

Hi, I am currently working on optimizing an element based on two sliders.

And i have 60 similar elements with different context each which i want to optimise the same as the first element.

For one element, its coming perfectly using both the solvers.

But when i connect it to gene pool of 60 + 60 = 120 gene count. its not solving properly. i left it for almost 15hr, still not solved.

which solver should use , Evolutionary or annealing?

in annealing solver , how to choose previous track solution?

is there any way to get it?

Note: Fit conditions are directly related to the gene pool. i hope there is no problem in fit condition.

Replies to This Discussion

Permalink Reply by s s solomon on February 18, 2014 at 3:19am

i have attached the pics of single element solution and overall solution.

I have marked the sample errors where it has gap.

Permalink Reply by Richard Schaffranek on February 18, 2014 at 3:27am

I am not entirely sure since the description of the problem is missing but:

lets assume you slider has two possible states 0 and 1 then with two sliders there are four possibilities:

00, 01, 10, 11

with three sliders 8:

000, 001, 010, 011, 100, 101, 110, 111

with 120 slider -> 2^120 = 1.329 * 10^36

http://www.wolframalpha.com/input/?i=+2%5E120

How do you combine the fitness function? Galapagos only takes 1 number as a fitness so this might be a problem as well, simple mass addition might not do the trick.

Permalink Reply by s s solomon on February 18, 2014 at 3:33am

Thanks Richard, Yeah you got the point. How can i solve for 60 different fitness values. everything passed as separate list.?

Permalink Reply by Richard Schaffranek on February 18, 2014 at 3:53am

You can't it only accepts one fitness value, end...

Again what is the fitness function? Galapagos can not find the best solution or a right solution it can only compare solutions and evaluate which one is better.So using Galapagos to find solutions without "error" is a bad use case, all solutions should be correct and than you can use Galapagos to find a fitter version.

As it looks from the images the question is if you need Galapagos to solve it... it looks like you could find a way how to construct the result.

Permalink Reply by s s solomon on February 18, 2014 at 4:15am

Thank you so much for your advice Richard, I will work more on it. this is one my 1st real life project with grasshopper.

RSS