how to divid a shape

Replies to This Discussion

Permalink Reply by Robert Vier on June 24, 2014 at 2:51

Hi,

you basically just ran it for four generations. If you increase the runtime, results get significantly better - 20 to 50 generations will yield a very good approximation already.

Also, for this problem, a low mutation rate and high elitism are beneficial - see attached!

The only thing with Octopus is that it does not have a viewport to show an arbitrary number of objectives of the solution space. I ll try to add an alternative viewport as a 2D chart which shows all goal dimensions of the solutions.

Best,

Robert

Attachments:

• octopustest.gh, 214 KB

Permalink Reply by dcjxb on June 30, 2014 at 4:06

Thank you for your answer,I have another question ,is the green cube that most close to the orgin the best solution? 

Permalink Reply by Robert Vier on June 30, 2014 at 8:08

I you are optimizing for 5 dimensions, the biggest, greenest cube nearest to the origin is the one which is most balanced between the objectives, an average trade-off. but this always depends on the solutions' distribution in the different objective dimensions. in that regard, for this problem, yes this would be one of the best trade-offs. BUT as you have 6 goal dimensions there is one missing - so in the viewport (for now) you dont really see a one-time-picture representation of the solution space.

you could bypass this by exporting the solutions to text files and import them to grasshopper with octopus.e - then make yourself a custom solution space graph just as the gene-display in octopus - with one axe for each objective. as said i plan to implement this into octopus soon.

best

Robert

Permalink Reply by dcjxb on July 1, 2014 at 22:36

Thank you very much.I'll try